Search

Handmade paper is a major carrier and restoration material of traditional Chinese ancient books, calligraphies, and paintings. In this study, we carried out a Raman spectroscopy analysis of 18 types of handmade paper samples. The main components of the handmade paper were cellulose and lignin, according to the wavenumber and Raman vibration assignment. We divided its Raman spectrum into eight subbands. Five machine learning models were employed: principal component analysis (PCA), partial least squares (PLS), support vector machine (SVM), k‐nearest neighbors (KNN), and random forest (RF). The Raman spectral data were normalized, and the fluorescence envelope was subtracted using the airPLS algorithm to obtain four types of data, raw, normalized, defluorescence, and fluorescence data. An RF variable importance analysis of data processing showed that data normalization eliminated the intensity differences of fluorescence signals caused by lignin, which contained important information of raw materials and papermaking technology, let alone the data defluorescence. The data processing also reduced the importance of the average variables in almost all spectral bands. Nevertheless, the data processing is worthwhile because it significantly improves the accuracy of machine learning, and the information loss does not affect the prediction. Using the machine learning models of PCA, PLS, and SVM combined with linear regression (LR), KNN, and RF, the classification and prediction of handmade paper samples were realized. For almost all processed data, including the fluorescence data, PCA‐LR had the highest classification and prediction accuracy (R2 = 1) in almost all spectral bands. PLS‐LR and SVM‐LR had the second‐highest accuracies (R2 = 0.4–0.9), whereas KNN and RF had the lowest accuracies (R2 = 0.1–0.4) for full band spectral data. Our results suggest that the abundant information contained in Raman spectroscopy combined with powerful machine learning models could inspire further studies on handmade paper and related cultural relics. [ABSTRACT FROM AUTHOR]

In recent years, frequency domain spectroscopy (FDS) is often used to evaluate oil paper insulation state in power transformer bushing. But it is still very difficult to evaluate the moisture content accurately and quickly. In order to solve this problem, this paper proposes an intelligent algorithm based on random forest regression (RFR) to construct an efficient evaluation method through segmented FDS curves. Furthermore, the characteristics of FDS curves were studied and the intelligent method was compared with support vector regression (SVR) and deep neural networks (DNN). The results show that the dielectric loss, the real part and imaginary part of complex capacitance all move upward with the moisture increasing, so they can be used as the input feature of the evaluation model; The moisture content evaluation accuracy of the RFR model in the whole frequency band is higher than that of SVR and DNN models; With the increase of lower cut off frequency (FDS test stop frequency), the FDS test time is greatly shortened, and the accuracy of the RFR model can still meet the evaluation requirements. Therefore, the data in a compromise frequency band can be used to evaluate the moisture content of oil paper insulation accurately and quickly. [ABSTRACT FROM AUTHOR]

This work presents machine learning (ML) assisted Fermi level prediction of solution‐processed ultra‐wide bandgap (UWB) amorphous gallium oxide (a‐Ga2Ox) which can significantly accelerate the fabrication of semiconducting UWB a‐Ga2Ox‐based material for future display application. Different models such as Kernel Ridge Regression (KRR), Support Vector Regression (SVR) and Random Forest Regression (RFR) were trained with empirical features, including experimental thickness, annealing temperature and environment during the solution‐processed UWB a‐Ga2Ox film fabrication. This work is a big step towards rapid and cost‐effective optimization method of fabricating UWB a‐Ga2Ox‐based devices. [ABSTRACT FROM AUTHOR]

Inversion of magnetic basement interfaces in basins is essential for interpreting potential field data and studying geothermal resource distribution, as well as basin formation and evolution. This paper introduces a novel method for inverting magnetic basement interfaces using a random forest regression (RFR) algorithm that combines potential field processing and machine learning techniques. The method creates magnetic base interface models and corresponding magnetic anomaly data through the random midpoint displacement method and magnetic interface finite element forward simulation. These anomalies are then processed using techniques such as directional transformations, analytical continuation, spatial derivatives, and fractional transformations. Feature attributes are extracted, and Gini importance is utilized to measure the contributions of feature factors, identify effective features, and enhance algorithm efficiency. The validity and practicality of the method are demonstrated through the analysis of both idealized and noisy models. The proposed machine learning‐based approach is more intelligent, efficient, and accurately represents the relief of magnetic base interfaces. When applied to magnetic survey data in the Datong Basin, it produced a reliable basin base model that aligns with known structural information, paving the way for further research in magnetic interface inversion. Key Points: A machine learning method based on random forest regression algorithm is used to recover the relief of the magnetic basement interface by using high‐precision magnetic dataApplying potential field processing and transformation to feature extraction for machine learningThe proposed technique intelligently and efficiently restores the 3D relief of the magnetic basement interface, and it has been successfully implemented in the Datong Basin's Tianzhen‐Yanggao sag [ABSTRACT FROM AUTHOR]

This paper illustrates the application of machine learning algorithms in predictive analytics for local governments using administrative data. The developed and tested machine learning predictive algorithm overcomes known limitations of the conventional ordinary least squares method. Such limitations include but not limited to imposed linearity, presumed causality with independent variables as presumed causes and dependent variables as presume result, likely high multicollinearity among features, and spatial autocorrelation. The study applies the algorithms to 311 non‐emergency service requests in the context of Miami‐Dade County. The algorithms are applied to predict the volume of 311 service requests and the community characteristics affecting the volume across Census tract neighborhoods. Four common families of algorithms and an ensemble of them are applied. They are random forest, support vector machines, lasso and elastic‐net regularized generalized linear models, and extreme gradient boosting. Two feature selection methods, namely Boruta and fscaret, are applied to identify the significant community characteristics. The results show that the machine learning algorithms capture spatial autocorrelation and clustering. The features generated by fscaret algorithms are parsimonious in predicting the 311 service request volume. [ABSTRACT FROM AUTHOR]

Spectral data describing product samples are typically composed of a large number of noisy and irrelevant wavelengths that tends to undermine the performance of multivariate predictive techniques. This paper proposes a two‐phase framework that integrates a preselection wavelength step oriented by wavelength clustering to a wrapper‐based strategy. The first phase performs a pruning process in the data that removes the less informative wavelengths relying on the spectral clustering, a technique deemed suitable to the Fourier transform infrared (FTIR) spectroscopy and near‐infrared (NIR) spectroscopy data at hand. The preselected wavelengths undergo a second phase of selection efforts based on the combination of different wavelength importance indices (i.e., Bhattacharyya distance, Chi‐square, ReliefF, and Gini) and classification techniques (i.e., support vector machine, k‐nearest neighbors, and random forest). When applied to 11 FTIR datasets from different domains, the recommended combination of importance index and classifier increased the average accuracy by 6.37% (from 0.863 to 0.918), while retaining average 3.84% of the original spectra. The framework also improved the selection process regarding computational time. This paper introduces a two‐phase framework merging wavelength preselection through clustering and a wrapper strategy. Initial spectral clustering eliminates less informative wavelengths. Subsequently, diverse wavelength importance indices and classification methods are integrated. Applied to 11 spectral datasets, the proposed combination (spectral clustering [SC]‐random forest [RF]+Gini [GI]) enhances average accuracy by 6.37%, retaining 3.84% of the original spectra, and reduces computational time. [ABSTRACT FROM AUTHOR]

In this paper, the historical network loss is calculated by using the cleaned historical data collected by the user electricity meter, the transformer gateway meter every 15 min and the gateway meter of the transformer substation every half an hour. Use machine learning technologies to analyze the loss of distribution network, reveal the changing law of distribution network loss and its influencing factors (electrical/non‐electrical parameters), and predict future distribution network loss. Pearson correlation, characteristics of the Random Forest important attributes and MIC (Maximal Information Coefficient) are adopted to analyze correlations between the power network loss and the electrical parameters such as unbalanced three‐phase, power factor, load rate, active power, reactive power together with atmospheric factors (temperature, humidity, and wind speed). The factors strongly related to network loss as the input of the multi‐parallel input multi‐step output stacking LSTM (Long Term Memory) prediction model to predict network loss. The simulation data in this paper comes from the actual data of 44 subareas of the 10 KV distribution network. The experimental results show that the data cleaning methods: the cluster‐based Local Outlier Factor algorithm and the improved Random Forest imputation algorithm are effective for enhancing the prediction accuracy. Especially the improved Random Forest algorithm is robust to all missing forms. Compared with the baseline model, the multi‐parallel input multi‐step output stacking LSTM prediction model has higher forecasting accuracy. [ABSTRACT FROM AUTHOR]

Summary: This study explores various machine learning and deep learning applications on financial data modelling, analysis and prediction processes. The main focus is to test the prediction accuracy of cryptocurrency hourly returns and to explore, analyse and showcase the various interpretability features of the ML models. The study considers the six most dominant cryptocurrencies in the market: Bitcoin, Ethereum, Binance Coin, Cardano, Ripple and Litecoin. The experimental settings explore the formation of the corresponding datasets from technical, fundamental and statistical analysis. The paper compares various existing and enhanced algorithms and explains their results, features and limitations. The algorithms include decision trees, random forests and ensemble methods, SVM, neural networks, single and multiple features N‐BEATS, ARIMA and Google AutoML. From experimental results, we see that predicting cryptocurrency returns is possible. However, prediction algorithms may not generalise for different assets and markets over long periods. There is no clear winner that satisfies all requirements, and the main choice of algorithm will be tied to the user needs and provided resources. [ABSTRACT FROM AUTHOR]

The ease with which mobile money is used to facilitate cross‐border payments presents a global threat to law enforcement in the fight against money laundering and terrorist financing. This paper aims to utilize machine learning classifiers to predict transactions flagged as a fraud in mobile money transfers. The data for this study were obtained from real‐time transactions that simulate a well‐known mobile transfer fraud scheme. Logistic regression is used as the baseline model and is compared with ensemble and gradient descent models. The results indicate that the logistic regression model still showed reasonable performance while not performing as well as the other models. Among all the measures, the random forest classifier exhibited outstanding performance. The amount of money transferred emerged as the top feature for predicting money laundering transactions in mobile money transfers. These findings suggest that further research is needed to enhance the logistic regression model, and the random forest classifier should be explored as a potential tool for law enforcement and financial institutions to detect money laundering activities in mobile money transfers. [ABSTRACT FROM AUTHOR]

In this paper, we survey the most recent advances in supervised machine learning (ML) and high‐dimensional models for time‐series forecasting. We consider both linear and nonlinear alternatives. Among the linear methods, we pay special attention to penalized regressions and ensemble of models. The nonlinear methods considered in the paper include shallow and deep neural networks, in their feedforward and recurrent versions, and tree‐based methods, such as random forests and boosted trees. We also consider ensemble and hybrid models by combining ingredients from different alternatives. Tests for superior predictive ability are briefly reviewed. Finally, we discuss application of ML in economics and finance and provide an illustration with high‐frequency financial data. [ABSTRACT FROM AUTHOR]

Summary: Summary Software‐Defined Networking (SDN) is an emerging network paradigm that has gained significant traction from many researchers to address the requirement of current data centers. Although central control is the major advantage of SDN, it is also a single point of failure if it is made unreachable by a Distributed Denial of Service (DDoS) attack. Despite the large number of traditional detection solutions that exist currently, DDoS attacks continue to grow in frequency, volume, and severity. This paper brings an analysis of the problem and suggests the implementation of four machine learning algorithms (SVM, MLP, Decision Tree, and Random Forest) with the purpose of classifying DDoS attacks in an SDN simulated environment (Mininet 2.2.2). With this goal, the DDoS attacks were simulated using the Scapy tool with a list of valid IPs, acquiring, as a result, the best accuracy with the Random Forest algorithm and the best processing time with the Decision Tree algorithm. Moreover, it is shown the most important features to classify DDoS attacks and some drawbacks in the implementation of a classifier to detect the three kinds of DDoS attacks discussed in this paper (controller attack, flow‐table attack, and bandwidth attack). [ABSTRACT FROM AUTHOR]

Stroke is the leading cause of permanent disability in adults, and it can cause permanent brain damage. According to the World Health Organization, 795 000 Americans experience a new or recurrent stroke each year. Early detection of medical disorders, for example, strokes, can minimize the disabling effects. Thus, in this paper, we consider various risk factors that contribute to the occurrence of stoke and machine learning algorithms, for example, the decision tree, random forest, and naive Bayes algorithms, on patient characteristics survey data to achieve high prediction accuracy. We also consider the semisupervised self‐training technique to predict the risk of stroke. We then consider the near‐miss undersampling technique, which can select only instances in larger classes with the smaller class instances. Experimental results demonstrate that the proposed method obtains an accuracy of approximately 98.83% at low cost, which is significantly higher and more reliable compared with the compared techniques. [ABSTRACT FROM AUTHOR]

AI and machine learning are increasingly often applied in the medical industry. The COVID‐19 epidemic will start to spread quickly over the planet around the start of 2020. At hospitals, there were more patients than there were beds. It was challenging for medical personnel to identify the patient who needed treatment right away. A machine learning approach is used to predict COVID‐19 pandemic patients at high risk. To provide input data and output results that execute the machine learning model on the backend, a straightforward Python Flask web application is employed. Here, the XGBoost algorithm, a supervised machine learning method, is applied. In order to predict high‐risk patients based on their current underlying health issues, the model uses patient characteristics as well as criteria like age, sex, health issues including diabetes, asthma, hypertension, and smoking, among others. The XGBoost model predicts the patient's severity with an accuracy of about 98% after data pre‐processing and training. The most important factors to the models are chosen to be age, diabetes, sex, and obesity. Patients and hospital personnel will benefit from this project's assistance in making timely choices and taking appropriate action. This will let medical personnel decide how much time and space to devote to the COVID‐19 high‐risk patients. providing a treatment that is both efficient and ideal. With this programme and the necessary patient data, hospitals may decide whether a patient need immediate care or not. [ABSTRACT FROM AUTHOR]

Geomorphology generally aims to describe and investigate the processes that lead to the formation of landscapes, while geochronology is needed to detect their timing and duration. Due to restrictions on exporting geological samples from Egypt, modern geoscientific studies in the Nile Delta lack the possibility of dating the investigated sediments and geological features by standard techniques such as OSL or AMS 14C; therefore, this study aims to validate a new approach using machine‐learning algorithms on portable X‐ray fluorescence (pXRF) data. Archaeologically dated sediments from the archaeological excavations of Buto (Tell el‐Fara'in; on‐site) that pXRF analyses have geochemically characterized serve as training data for running and comparing Neural Nets, Random Forests, and single‐decision trees. The established pXRF fingerprints are transferred via machine‐learning algorithms to set up a chronology for undated sediments from sediment cores (i.e., the test data) of the nearby surroundings (off‐site). Neural Nets and Random Forests work fine in dating sediments and deliver the best classification results compared with single‐decision trees, which struggle with outliers and tend to overfit the training data. Furthermore, Random Forests can be modeled faster and are easier to understand than the complex, less transparent Neural Nets. Therefore, Random Forests provide the best algorithm for studies like this. Furthermore, river features east of Kom el‐Gir are dated to pre‐Ptolemaic times (before 332 B.C.) when Kom el‐Gir had possibly not yet been settled. The research in this paper shows the success of close interactions from various scientific disciplines (Geoinformatics, Physical Geography, Archaeology, Ancient History) to decipher landscape evolution in the long‐term‐settled Nile Delta's environs using machine learning. With the approach's design and the possibility of integrating many other geographical/sedimentological methods, this study demonstrates the potential of the methodological approach to be applied in other geoscientific fields. Highlights: Neural Nets and Random Forest yield excellent results in classifying sedimentsC5.0 single‐decision trees showed substantial deficits in this context [ABSTRACT FROM AUTHOR]

With the continuous development of machine learning and the increasing complexity of financial data analysis, it is more popular to use models in the field of machine learning to solve the hot and difficult problems in the financial industry. To improve the effectiveness of stock trend prediction and solve the problems in time series data processing, this paper combines the fuzzy affiliation function with stock‐related technical indicators to obtain nominal data that can widely reflect the constituent stocks in the case of time series changes by analysing the S&P 500 index. Meanwhile, in order to optimise the current machine learning algorithm in which the setting and adjustment of hyperparameters rely too much on empirical knowledge, this paper combines the deep forest model to train the stock data separately. The experimental results show that (1) the accuracy of the extreme random forest and the accuracy of the multi‐grain cascade forest are both higher than that of the gated recurrent unit (GRU) model when the un‐fuzzy index‐adjusted dataset is used as features for input, (2) the accuracy of the extreme random forest and the accuracy of the multigranular cascade forest are improved by using the fuzzy index‐adjusted dataset as features for input, (3) the accuracy of the fuzzy index‐adjusted dataset as features for inputting the extreme random forest is improved by 18.89% compared to that of the un‐fuzzy index‐adjusted dataset as features for inputting the extreme random forest and (4) the average accuracy of the fuzzy index‐adjusted dataset as features for inputting multi‐grain cascade forest increased by 5.67%. [ABSTRACT FROM AUTHOR]

Cancer has been the leading cause of death among Japanese since 1981, and many people die from it every year worldwide. While various measures have been taken to reduce the mortality rate of cancer, circulating tumor cells (CTCs) in the blood have been attracting attention in recent years. In the past, CTCs were detected by visual inspection by a physician or by an expensive machine, but these methods required much effort by the physician and required only EpCAM‐expressing cells to be detected. In addition, detection by image processing has been used, but it has the problem that the area of interest is only a part of the area and there are many false positives. In this paper, we propose a two‐step classification method that focuses on the shape and surface of cells. In the proposed method, multiple shape and surface features are obtained for four types of cells in blood images: Clusters, CTCs, Normal Cells, and Vertical Cells. Based on the features, cells are classified using a two‐step Random Forest and their accuracy is evaluated. Furthermore, the effectiveness of the proposed method is demonstrated by comparing it with conventional methods. [ABSTRACT FROM AUTHOR]

Increasing impacts of wildfires on Western US air quality highlights the need for forecasts of smoke emissions based on dynamic modeled wildfires. This work utilizes knowledge of weather, fuels, topography, and firefighting, combined with machine learning and other statistical methods, to generate 1‐ and 2‐day forecasts of fire radiative energy (FRE). The models are trained on data covering 2019 and 2021 and evaluated on data for 2020. For the 1‐day (2‐day) forecasts, the random forest model shows the most skill, explaining 48% (25%) of the variance in observed daily FRE when trained on all available predictors compared to the 2% (<0%) of variance explained by persistence for the extreme fire year of 2020. The random forest model also shows improved skill in forecasting day‐to‐day increases and decreases in FRE, with 28% (39%) of observed increase (decrease) days predicted, and increase (decrease) days are identified with 62% (60%) accuracy. Error in the random forest increases with FRE, and the random forest tends toward persistence under severe fire weather. Sensitivity analysis shows that near‐surface weather and the latest observed FRE contribute the most to the skill of the model. When the random forest model was trained on subsets of the training data produced by agencies (e.g., the Canadian or US Forest Services), comparable if not better performance was achieved (1‐day R2 = 0.39–0.48, 2‐day R2 = 0.13–0.34). FRE is used to compute emissions, so these results demonstrate potential for improved fire emissions forecasts for air quality models. Plain Language Summary: Increasing wildfire smoke is undoing decades of air quality progress, yet air quality forecasts often miss the most intense smoke events. This is because forecasted smoke is released at constant rates whereas the rate of smoke release from real wildfires varies in time. In this work we teach a machine learning algorithm to predict the daily change in fire heat output, a quantity that is used to calculate wildfire emissions. The machine learning algorithm uses information regarding weather, fuel moisture and amount, and firefighting efforts to make its predictions. We also tried to predict the daily change in fire heat output using only weather information but found the machine learning method to be more successful. Many federal agencies have their own ways of tracking fire weather and fuel moisture, and in this paper, we show that we can apply machine learning to the data from any of several agencies and get the same level of forecasting skill. Key Points: Random forest models trained on weather, fuel, and firefighting data surpass persistence and weather‐based methods to predict fire energyRandom forest models beat persistence across states, for most days in the 2020 fire season, and across levels of fire severityFire weather and latest fire energy predictors add most skill to the random forest; models using agency weather data perform similarly [ABSTRACT FROM AUTHOR]

This methodology is the analysis backbone at Optimizely, one of the largest third-party experimentation platform vendors.[14] See also Deng et al.[15] for a Bayesian approach to continuously monitor OCEs using Bayes factors. 26 Deng A, Xu Y, Kohavi R, Walker T, Improving the sensitivity of online controlled experiments by utilizing pre-experiment data. Most often, hyperparameter tuning is done via grid search (i.e., factorial designs) but this problem is perfectly suited for response surface methodology to both efficiently determine which hyperparameters are important (via screening experiments) and then to determine the optimal configuration of the important ones via response surface experiments and second order models. Keywords: design and analysis of experiments; hyperparameter tuning; simulation design EN design and analysis of experiments hyperparameter tuning simulation design 1107 1109 3 05/09/23 20230601 NES 230601 I would first like to congratulate Dr. Anderson-Cook and Dr. Lu for an important, insightful, and timely paper. [Extracted from the article]

This paper presents the machine learning (ML) algorithm to predict the thermal performance of closed‐loop thermosyphon (CLT). The experimentation is carried out on the acetone‐charged CLT at different test conditions such as heat inputs, filling ratios, and adiabatic lengths. The test data is used to calculate the performance parameters such as thermal resistance, heat transfer coefficient, and effectiveness of the system. Based on the experimental dataset, the ML algorithms are developed to predict the performance parameters of the CLT system. The ML algorithms such as linear regression, decision tree (DT), random forest (RF), and lasso regression are used for the development of the prediction model. The hyperparameters are well‐tuned and optimized. The prediction measuring parameters (mean absolute error, R2) are analyzed carefully. It is noticed that the DT model outperformed the prediction of the other used models. The R2 score of the DT model was 98.504; whereas, the R2 scores of the RF model and linear regression model were about 94.76 and 92.17, respectively. This study will become a roadmap to the ML approach in the thermosyphon system. [ABSTRACT FROM AUTHOR]

Concrete which is the most commercialized construction material and thus it plays a key role in this era of development and hence its evolution is of utmost importance and therefore the evolution of the quality of concrete to that of its highly evolved type namely, ultra‐high performance concrete (UHPC) is undeniably the boon to this sector. Though, the correlations between the technical characteristics of UHPC and the composition of its mixture are complicated, nonlinear, and complex to characterize using standard statistical techniques. This paper is intended to use both deep neural network and ensemble machine learning algorithms namely gradient boosting, extreme gradient boosting, random forest regressor, extra tree regressor, and voting regressor trained with an 810 UHPC mixture collections with 15 input variables to predict its compressive strength. After adjusting a regression model, the prediction efficiency and generalization ability of the developed models are validated using a number of performance parameters. It was established that all employed models performed better at forecasting result, the extra tree regressor was the most accurate followed by extreme gradient boosting. [ABSTRACT FROM AUTHOR]

This paper aims to study the predictive power of market herding effect for market volatility. We extend the widely used four types of linear HAR models by incorporating the market herding index and further introduce a nonlinear forecasting mechanism by using two machine learning algorithms—support vector regression (SVR) and random forest (RF). All the methods are applied to the Chinese stock market. We evaluate their in‐sample and out‐of‐sample performances for various prediction step lengths. The results show that according to the RMSE and MAE, for the daily volatility prediction, the extra herding factor makes no significant contribution, and the linear model has no significant difference with the algorithms as well, whereas with the increase of step length, the herding factor shows a significant power in the weekly, monthly, and quarterly prediction, and meanwhile, the RF algorithm outperforms the other methods. However, for the QLIKE, the herding factor always significantly improves the predictions for all kinds of step lengths and methods, and the SVR algorithm with the herding factor is dominant. Overall, the herding factor helps forecast the volatility, and the extent to which the precision is improved takes on an inverted "U" shape with the increase in step length. The two algorithms are superior to the linear model, especially for the longer‐term prediction, and the herding factor can help enlarge this advantage, which implies a nonlinear relationship between the herding and volatility. Besides, the extra liquidity factor plays no significant role in the volatility prediction, but the continuous variation (CV), jump variation (JV), and leverage effect are helpful in most cases. [ABSTRACT FROM AUTHOR]

A phoneme is perceptually the smallest distinct sound unit distinguished among words in a particular language. Every language has its own set of phonemes, and all the words are ordered sequences of phonemes. Therefore, phoneme recognition is essential to automatic speech recognition (ASR) systems. Phonemes of a language can be classified together using a single machine learning (ML) model through the direct classification (also known as the baseline or flat classification) approach. However, it is observed that the performance of such phoneme recognition degrades with the increase in the number of phoneme classes. The challenge is pronounced in languages with a larger number of phoneme classes, like Hindi, which has 48 phonemes. In this paper, we propose a speaker‐independent hierarchical classification approach for 33 Hindi‐Devanagari consonants/phonemes using cepstral features with ML techniques like support vector machine (SVM), random forest (RF) and fully connected deep neural network (DNN). In this hierarchical approach, a given phoneme is classified into successive subgroups until the particular phoneme class is identified. To perform the classification task, a binary or multi‐class classifier is invoked for each internal (non‐leaf) node in the hierarchy tree. Our model identified pairs of Optimal Feature Sets (based on mutual information) and the best suitable ML classifier for each internal decision node in the hierarchy using 10‐fold cross‐validation to help in efficient classification. Our proposed hierarchical model leads to better accuracy and 57% improved performance for phoneme recognition compared with the non‐hierarchical, that is, the direct classification approach. [ABSTRACT FROM AUTHOR]

Ever since Ether is launched as a digital currency, its rise has been rapid. It is currently the second most valuable digital currency in the world. There are more than 1 million transactions happening on the Ethereum network every day, and this number is expected to continue to increase. Due to the increasing number of transactions, fraudulent transactions have also increased, which has resulted in a large amount of money being lost and has also destroyed the livelihoods of many individuals. Due to their similarity to valid transactions, it is extremely difficult to distinguish between them. Additionally, Ethereum's pseudo‐anonymity adds to the difficulty of identifying the parties involved. Since there are millions of transactions every day, it would be difficult to manually verify each one. Therefore, a mechanism for validating these transactions is needed. In this context, this paper proposes a novel approach to detecting fraudulent accounts associated with these transactions by implementing machine learning algorithms among the given set of transactions. We propose a framework for creating a stacking classifier by combining several standalone classification algorithms and creating a meta‐learner based on the output of each base algorithm. The algorithms include Logistic Regression, Naive Bayes, Decision Trees, Random Forests, AdaBoosts, KNNs, SVMs, and Gradient Boosts. As a result of combining these algorithms, a powerful classifier with the ability to detect fraudulent transactions. A variety of machine learning models were trained and evaluated on the test set using various metrics. Based on the results of the individual algorithm the Random Forest algorithm achieved the highest accuracy of 95.47%, followed by Gradient Boosting at 94.61% which is an ensemble algorithm using the boosting technique. The Stacking classifier that combines Multinomial Naive Bayes and Random Forest as the base learners and logistic regression as the Meta learner achieved the highest accuracy of 97.18% with an F1 score of 97.02%. Based on the results of all the stacking models developed, it is concluded that algorithms tend to perform better when combined properly. When compared to the other approaches, the proposed approach has outperformed the others, making it feasible in the real world to detect fraudulent transactions. [ABSTRACT FROM AUTHOR]

Predicting enantioselectivities in asymmetric catalytic reactions through chemometric approach is a challenging area. In this paper, quantitative structure−selectivity relationship (QSSR) models were successfully developed for thiol addition to N‐acylimines catalyzed by chiral phosphoric acids. Ten Dragon molecular descriptors calculated from thiols and phosphoric acid catalysts were used to correlate with 1075 enantioselectivities ΔΔG. Machine learning algorithms, support vector machine (SVM) and random forest (RF), were adopted to build QSSRs after descriptor subset selection based on the multiple linear regression technique. The SVM and RF models, respectively, have coefficient of determination R2 of 0.925 and 0.944 for the training set, and of 0.837 and 0.851 for the test set. Both SVM and RF models were subjected to internal and external validation and estimation in prediction ability and systematic error. [ABSTRACT FROM AUTHOR]

Species distribution modeling (SDM) is widely used in ecology and conservation. Currently, the most available data for SDM are species presence‐only records (available through digital databases). There have been many studies comparing the performance of alternative algorithms for modeling presence‐only data. Among these, a 2006 paper from Elith and colleagues has been particularly influential in the field, partly because they used several novel methods (at the time) on a global data set that included independent presence–absence records for model evaluation. Since its publication, some of the algorithms have been further developed and new ones have emerged. In this paper, we explore patterns in predictive performance across methods, by reanalyzing the same data set (225 species from six different regions) using updated modeling knowledge and practices. We apply well‐established methods such as generalized additive models and MaxEnt, alongside others that have received attention more recently, including regularized regressions, point‐process weighted regressions, random forests, XGBoost, support vector machines, and the ensemble modeling framework biomod. All the methods we use include background samples (a sample of environments in the landscape) for model fitting. We explore impacts of using weights on the presence and background points in model fitting. We introduce new ways of evaluating models fitted to these data, using the area under the precision‐recall gain curve, and focusing on the rank of results. We find that the way models are fitted matters. The top method was an ensemble of tuned individual models. In contrast, ensembles built using the biomod framework with default parameters performed no better than single moderate performing models. Similarly, the second top performing method was a random forest parameterized to deal with many background samples (contrasted to relatively few presence records), which substantially outperformed other random forest implementations. We find that, in general, nonparametric techniques with the capability of controlling for model complexity outperformed traditional regression methods, with MaxEnt and boosted regression trees still among the top performing models. All the data and code with working examples are provided to make this study fully reproducible. [ABSTRACT FROM AUTHOR]

Background Aims Methods & Procedures Outcomes & Results Conclusions & Implications WHAT THIS PAPER ADDS What is already known on the subject What this study adds to the existing knowledge What are the potential or clinical implications of this work? Identifying language disorders earlier can help children receive the support needed to improve developmental outcomes and quality of life. Despite the prevalence and impacts of persistent language disorder, there are surprisingly no robust predictor tools available. This makes it difficult for researchers to recruit young children into early intervention trials, which in turn impedes advances in providing effective early interventions to children who need it.To validate externally a predictor set of six variables previously identified to be predictive of language at 11 years of age, using data from the Longitudinal Study of Australian Children (LSAC) birth cohort. Also, to examine whether additional LSAC variables arose as predictive of language outcome.A total of 5107 children were recruited to LSAC with developmental measures collected from 0 to 3 years. At 11–12 years, children completed the Clinical Evaluation of Language Fundamentals, 4th Edition, Recalling Sentences subtest. We used SuperLearner to estimate the accuracy of six previously identified parent‐reported variables from ages 2–3 years in predicting low language (sentence recall score ≥ 1.5 SD below the mean) at 11–12 years. Random forests were used to identify any additional variables predictive of language outcome.Complete data were available for 523 participants (52.20% girls), 27 (5.16%) of whom had a low language score. The six predictors yielded fair accuracy: 78% sensitivity (95% confidence interval (CI) = [58, 91]) and 71% specificity (95% CI = [67, 75]). These predictors relate to sentence complexity, vocabulary and behaviour. The random forests analysis identified similar predictors.We identified an ultra‐short set of variables that predicts 11–12‐year language outcome with ‘fair’ accuracy. In one of few replication studies of this scale in the field, these methods have now been conducted across two population‐based cohorts, with consistent results. An imminent practical implication of these findings is using these predictors to aid recruitment into early language intervention studies. Future research can continue to refine the accuracy of early predictors to work towards earlier identification in a clinical context. There are no robust predictor sets of child language disorder despite its prevalence and far‐reaching impacts. A previous study identified six variables collected at age 2–3 years that predicted 11–12‐year language with 75% sensitivity and 81% specificity, which warranted replication in a separate cohort.There are no robust predictor sets of child language disorder despite its prevalence and far‐reaching impacts. A previous study identified six variables collected at age 2–3 years that predicted 11–12‐year language with 75% sensitivity and 81% specificity, which warranted replication in a separate cohort. We used machine learning methods to identify a set of six questions asked at age 2–3 years with ≥ 71% sensitivity and specificity for predicting low language outcome at 11–12 years, now showing consistent results across two large‐scale population‐based cohort studies.We used machine learning methods to identify a set of six questions asked at age 2–3 years with ≥ 71% sensitivity and specificity for predicting low language outcome at 11–12 years, now showing consistent results across two large‐scale population‐based cohort studies. This predictor set is more accurate than existing feasible methods and can be translated into a low‐resource and time‐efficient recruitment tool for early language intervention studies, leading to improved clinical service provision for young children likely to have persisting language difficulties.This predictor set is more accurate than existing feasible methods and can be translated into a low‐resource and time‐efficient recruitment tool for early language intervention studies, leading to improved clinical service provision for young children likely to have persisting language difficulties. [ABSTRACT FROM AUTHOR]

With an increasingly urbanized world, there is an urgent need to examine how cities may evolve and achieve sustainability. This paper systematically looks at the Greater Bay Area (GBA) and the Poyang Lake Region (PLR) in China to examine the spatial processes for insights into cities and urbanization, balancing the environmental and socio‐economic dimensions. A total of 226 805 cells are analyzed to unveil the relationship between sustainability changes in 2015–2019 period and urban form indicators, considering sociodemographic variables, geographical features, and city size as control variables. Two tree‐based machine learning models (Random Forest and XGBoost) are developed. This study provides evidence that a monocentric urban form and a high share of small activity clusters are not good for sustainability. For each urban form indicator, there is a non‐linear relationship with sustainability. The results of the machine learning models reconfirm the sustainability benefits of having a strong second activity cluster comparable to the largest one. When planning cities, some forms of land use buffering are desirable. There is also support for developing relatively large activity nodes and promoting compactness in urban form. Beyond urban form characteristics, the levels of urbanization, economic development, and population are still highly relevant. [ABSTRACT FROM AUTHOR]

The development of fragility curves is an important step in seismic risk assessment within the scope of performance‐based earthquake engineering. The goal of this work is to use machine learning methods (regression‐based tools) to forecast the large‐span precast parking structural responses and fragility curves. This study proposes five predicting models based on machine learning to evaluate the seismic performance of the large‐span precast parking structures, including: neural networks, genetic algorithm‐based neural networks, support vector machine, decision tree and random forest. A database that includes 453 numerical synthetic results was used to train and test the machine learning models. The seismic performance of large‐span precast parking structures were predicted using the constructed machine learning models. Finally, the sensitivity analysis of input parameters was conducted. From this paper we can conclude that: (1) the genetic optimization‐based neural networks' predicting model has the most accurate predictive ability for seismic fragility estimation and (2) the structural responses and the fragility curves of parking structures are related to the differences of the stiffness of the connectors and the number of floors, of which the stiffness of the connectors should be given special attention. [ABSTRACT FROM AUTHOR]

Cost‐sensitive ensemble learning as a combination of two approaches, ensemble learning and cost‐sensitive learning, enables generation of cost‐sensitive tree‐based ensemble models using the cost‐sensitive decision tree (CSDT) learning algorithm. In general, tree‐based models characterize nice graphical representation that can explain a model's decision‐making process. However, the depth of the tree and the number of base models in the ensemble can be a limiting factor in comprehending the model's decision for each sample. The CSDT models are widely used in finance (e.g., credit scoring and fraud detection) but lack effective explanation methods. We previously addressed this gap with cost‐sensitive tree Shapley Additive Explanation Method (CSTreeSHAP), a cost‐sensitive tree explanation method for the single‐tree CSDT model. Here, we extend the introduced methodology to cost‐sensitive ensemble models, particularly cost‐sensitive random forest models. The paper details the theoretical foundation and implementation details of CSTreeSHAP for both single CSDT and ensemble models. The usefulness of the proposed method is demonstrated by providing explanations for single and ensemble CSDT models trained on well‐known benchmark credit scoring datasets. Finally, we apply our methodology and analyze the stability of explanations for those models compared to the cost‐insensitive tree‐based models. Our analysis reveals statistically significant differences between SHAP values despite seemingly similar global feature importance plots of the models. This highlights the value of our methodology as a comprehensive tool for explaining CSDT models. [ABSTRACT FROM AUTHOR]

Software effort estimation is a vital process in the software industry for successfully administering 5Ds of the software development life cycle (SDLC). The 5Ds stand for demand, development, direction, deployment, and designated cost of the software. Software development effort estimation (SDEE) is an effort prediction mechanism to calculate the effort for the development of the software product in order to minimize the challenges in the software field. Academics and practitioners are striving to identify which machine learning estimation technique yields more accurate results based on evaluation metrics, datasets, and other pertinent aspects. The feature selection techniques impact accuracy by selecting the main and relevant features in the dataset and eliminating the redundant and irrelevant features in the dataset. To achieve accurate estimations, the paper utilizes feature selection algorithms, along with various machine learning techniques, which predict the desired effort and the performance of the model has been measured in terms of prediction accuracy, R2 value, relative error, and mean absolute error. The datasets China and Maxwell are trained with the relevant features by applying feature selection algorithms, and estimation techniques are applied to predict the effort. The performance is compared with the regression models and feature selection techniques utilized by many authors previously. The result of the proposed methodology significantly gives the best performance with the combination of feature selection and estimation models than all regression models when applied alone, to both datasets. From the results, it is perceptible that random forest is performing well with the feature selection techniques and obtains the highest prediction accuracy of 99.33% with the China and 89.47% with the Maxwell datasets. [ABSTRACT FROM AUTHOR]

A machine‐learning classifier for radiation waveforms of negative return strokes (RSs) is built and tested based on the Random Forest classifier using a large data set consisting of 14,898 negative RSs and 159,277 intracloud (IC) pulses with 3‐D location information. Eleven simple parameters including three parameters related with pulse characteristics and eight parameters related with the relative strength of pulses are defined to build the classifier. Two parameters for the evaluation of the classifier performance are also defined, including the classification accuracy, which is the percentage of true RSs in all classified RSs, and the identification efficiency, which is the percentage of correctly classified RSs in all true RSs. The tradeoff between the accuracy and the efficiency is examined and simple methods to tune the tradeoff are developed. The classifier achieved the best overall performance with an accuracy of 98.84% and an efficiency of 98.81%. With the same technique, the classifier for positive RSs is also built and tested using a data set consisting of 8,700 positive RSs. The classifier has an accuracy of 99.04% and an efficiency of 98.37%. By examining misclassified waveforms, we show evidence that some RSs and IC discharges produce special radiation waveforms that are almost impossible to correctly classify without 3‐D location information, resulting in a fundamental difficulty to achieve very high accuracy and efficiency in the classification of lightning radiation waveforms. Plain Language Summary: Lightning location systems are required to classify return strokes (RSs) from intracloud (IC) discharges accurately and efficiently because the RS is the main discharge component that poses direct threats to the human society. In this paper, we report a machine‐learning classifier for negative RSs built using a large data set with accurate 3‐D location information. The classifier has an accuracy of 98.84% (98.84% of classified RSs are correct classifications) and an efficiency of 98.81% (98.81% of RSs can be correctly classified). With the same technique, we also built a classifier for positive RSs with similarly high accuracy and efficiency. Our classifiers only require some simple waveform parameters so the same technique can be used by various lightning location systems relatively easily. A sample Python script to use the classifier is provided and readers are encouraged to test the classifier using their own data set. We also demonstrate that some RSs and IC discharges produce abnormal waveforms, so 100% accuracy or efficiency is fundamentally difficult to realize using only waveform information. Key Points: A machine‐learning classifier for negative return strokes (RSs) is built using a large data set with 3‐D location informationBoth an accuracy and an efficiency of about 98.8% are achieved and the accuracy‐efficiency tradeoff can be easily controlledSome RSs and intracloud discharges produce special waveforms that are fundamentally difficult to classify without 3‐D location results [ABSTRACT FROM AUTHOR]

Modeling hydrological processes for managing the available water resources effectively is often complex due to the existence of high nonlinearity, and the associated prediction uncertainty mainly arising from model inputs, parameters, and structure. Despite several attempts to quantify the model prediction uncertainty, reducing the same for improving the reliability of models is indispensable for their wider acceptance. This paper presents a novel modeling framework for minimizing the prediction uncertainty in the streamflow simulation of the conceptual hydrological model (HBV) by integrating with the Bayesian‐based Particle Filter technique (PF) and machine learning algorithm (Random Forest algorithm, RF). Initially, the streamflow prediction interval (PI) is derived from the stochastically estimated parameters of the HBV model through the PF technique (HBV‐PF model). As the HBV‐PF model quantifies only parametric uncertainty, the RF algorithm was employed (HBV‐PF‐RF model) for further minimizing the prediction uncertainty by inherently taking care of different sources of uncertainty. The RF algorithm inherently combines the physics of the hydrological system (i.e., process‐based variables) with machine learning‐based approach to minimize the overall prediction uncertainty. The proposed framework was analyzed on Nepal and India's Sunkoshi and Beas River basins, through several statistical performance indices for assessing the accuracy and uncertainty of the model prediction. The framework was observed to be consistently improving the model performance minimizing the uncertainty in both watersheds. Therefore, the proposed framework can be considered to be more reliable in improving the prediction capability of hydrological models. Key Points: Development of a physics‐aware machine learning based hydrological model for streamflow simulationThe proposed framework characterizes the model prediction uncertainty, accounting different sources of uncertainty implicit/explicitlySignificant reduction in overall prediction uncertainty is achieved [ABSTRACT FROM AUTHOR]

Summary: This article provides an overview of supervised machine learning (ML) with a focus on applications in banking. The supervised ML techniques covered include bagging (random forest), boosting (gradient boosting machine) and neural networks. We begin with an introduction to ML tasks and techniques. This is followed by a description of tree‐based ensemble algorithms, including bagging with random forest and boosting with gradient boosting machines, as well as feedforward neural networks. We then provide an extensive discussion of hyper‐parameter optimisation techniques. Interpretability of ML results is an important topic in banking and other regulated industries, and it is also covered in some depth. The paper concludes with a comparison of the features of different ML algorithms and a discussion of their use in practice. An application from credit risk modelling in banking is used throughout the paper to illustrate the techniques and interpret the results of the algorithms. [ABSTRACT FROM AUTHOR]

The reliance on wireless services to exchange critical data is associated with various threats and attacks, which must be mitigated to ensure integrity and security of those wireless services. Posing a serious challenge to wireless systems, jamming is among these attacks. In order to mitigate jamming, directive antennas are used to minimize the signals that are received from the jammer, while maximizing the received legitimate signal from the authorized transmitter. In this paper, we propose a machine learning‐based anti‐jamming framework to provide a spatially dynamic and instantaneous anti‐jamming performance that is achieved at the physical layer. The proposed framework incorporates a dataset that can be deployed in the hardware of a receiver with a massive Multiple‐Inputs Multiple‐Outputs–(MIMO) antenna. Our extensive performance evaluation results demonstrate the effective performance of the proposed framework in preserving integrity of a massive–MIMO communication system despite the presence of a hostile jammer. Particularly, due to the tabular nature of the generated dataset, tree‐based random forest models achieved the best performance with a signal‐to‐interference‐plus‐noise ratio accuracy of 92%$$ 92\% $$ and fast anti‐jamming response in around 1.66 s under sever jamming conditions. [ABSTRACT FROM AUTHOR]

The long‐term monitoring of gross primary production (GPP) is crucial to the assessment of the carbon cycle of terrestrial ecosystems. In this study, a well‐known machine learning model (random forest, RF) is established to reconstruct the global GPP data set named ECGC_GPP. The model distinguished nine functional plant types, including C3 and C4 crops, using eddy fluxes, meteorological variables, and leaf area index (LAI) as training data of RF model. Based on ERA5_Land and the corrected GEOV2 data, global monthly GPP data set at a 0.05° resolution from 1999 to 2019 was estimated. The results showed that the RF model could explain 74.81% of the monthly variation of GPP in the testing data set, of which the average contribution of LAI reached 41.73%. The average annual and standard deviation of GPP during 1999–2019 were 117.14 ± 1.51 Pg C yr−1, with an upward trend of 0.21 Pg C yr−2 (p < 0.01). By using the plant functional type classification, the underestimation of cropland is improved. Therefore, ECGC_GPP provides reasonable global spatial pattern and long‐term trend of annual GPP. Plain Language Summary: Accurate estimation of gross primary production (GPP) is critical for understanding the terrestrial ecosystem carbon cycle. There are a variety of GPP data sets based on different methods, but huge differences validated by the GPP measured values of flux observation towers still exist. At present, a large amount of GPP measured data provides us with the opportunity to use machine learning models to estimate global GPP. This paper presents a new global GPP data set (ECGC_GPP) with 0.05° and monthly scales from 1999 to 2019. This GPP data set is based on random forest model and driven by remote sensing data from GEOV2 and climate data from ERA5_Land. In ECGC_GPP, site‐level training models are constructed based on plant functional types (especially C3 and C4 crops) to improve accuracy. All these improvements are aimed at improving the lack of interannual fluctuations and the underestimation of cropland in current machine learning data set. Key Points: The accuracy of gross primary production (GPP) estimation can be improved by distinguishing plant functional types, especially for C3 and C4 cropsSignificant increasing trend is found in this random forest‐based data setLeaf area index plays a leading role in both the average state and long‐term trend of GPP [ABSTRACT FROM AUTHOR]

Breast cancer is one of the leading causes of death among women worldwide. In most cases, the misinterpretation of medical diagnosis plays a vital role in increased fatality rates due to breast cancer. Breast cancer can be diagnosed by classifying tumors. There are two different types of tumors, such as malignant and benign tumors. Identifying the type of tumor is a tedious task, even for experts. Hence, an automated diagnosis is necessary. The role of machine learning in medical diagnosis is eminent as it provides more accurate results in classifying and predicting diseases. In this paper, we propose a deep ensemble network (DEN) method for classifying and predicting breast cancer. This method uses a stacked convolutional neural network, artificial neural network and recurrent neural network as the base classifiers in the ensemble. The random forest algorithm is used as the meta‐learner for providing the final prediction. Experimental results show that the proposed DEN technique outperforms all the existing approaches in terms of accuracy, sensitivity, specificity, F‐score and area under the curve (AUC) measures. The analysis of variance test proves that the proposed DEN model is statistically more significant than the other existing classification models; thus, the proposed approach may aid in the early detection and diagnosis of breast cancer in women, hence aiding in the development of early treatment techniques to increase survival rate. [ABSTRACT FROM AUTHOR]

Agriculture plays an important role in Senegalese economy and annual early warning predictions of crop yields are highly relevant in the context of climate change. In this study, we used three main machine learning methods (support vector machine, random forest, neural network) and one multiple linear regression method, namely Least Absolute Shrinkage and Selection Operator (LASSO), to predict yields of the main food staple crops (peanut, maize, millet and sorghum) in 24 departments of Senegal. Three combination of predictors (climate data, vegetation data or a combination of both) are used to compare the respective contribution of statistical methods and inputs in the predictive skill. Our results showed that the combination of climate and vegetation with the machine learning methods gives the best performance. The best prediction skill is obtained for peanut yield likely due to its high sensitivity to interannual climate variability. Although more research is needed to integrate the results of this study into an operational framework, this paper provides evidence of the promising performance machine learning methods. The development and operationalization of such prediction and their integration into operational early warning systems could increase resilience of Senegal to climate change and contribute to food security. [ABSTRACT FROM AUTHOR]

The evergrowing diversity of encrypted and anonymous network traffic makes network management more formidable to manage the network traffic. An intelligent system is essential to analyse and identify network traffic accurately. Network management needs such techniques to improve the Quality of Service and ensure the flow of secure network traffic. However, due to the usage of non‐standard ports and encryption of data payloads, the classical port‐based and payload‐based classification techniques fail to classify the secured network traffic. To solve the above‐mentioned problems, this paper proposed an effective deep learning‐based framework employed with flow‐time‐based features to predict heterogeneous secure network traffic best. The state‐of‐the‐art machine learning strategies (C4.5, random forest, and K‐nearest neighbour) are investigated for comparison. The proposed 1D‐CNN model achieved higher accuracy in classifying the heterogeneous secure network traffic. In the next step, the proposed deep learning model characterises the major categories (virtual private network traffic, the onion router network traffic, and plain encrypted network traffic) into several application types. The experimental results show the effectiveness and feasibility of the proposed deep learning framework, which yields improved predictive power compared to the state‐of‐the‐art machine learning techniques employed for secure network traffic analysis. [ABSTRACT FROM AUTHOR]

Occurrences of extreme events, especially floods, have become more frequent and severe in the recent past due to the global impacts of climate change. In this context, possibilities for generating a near‐accurate streamflow forecast at higher lead times, which could be utilized for developing a reliable flood warning system to minimize the effects of extreme events, are highly important. This paper aims to investigate the potential of a novel hybrid modeling framework that couples the random forest algorithm, particle filter, and the HBV model for improving the overall accuracy of forecasts at higher lead times through the dynamic error correction schematic. The new framework simulates an ensemble of streamflow for estimating uncertainty associated with the predictions and is applied across two snow‐fed Himalayan rivers: the Beas River in India and the Sunkoshi River in Nepal. Several statistical indices along with graphical performance indicators were used for assessing the accuracy of the model performance and associated uncertainty. The modeling framework achieved the Nash Sutcliffe Efficiency of 0.94 and 0.98 in calibration and 0.95 and 0.99 in validation for the Beas and Sunkoshi river basin respectively for a 7‐day ahead forecast. Thus, the proposed framework can be considered as a promising tool having reasonably good performance in forecasting streamflow at a higher lead time. Key Points: Hybrid hydrological model integrates process‐based model with machine learning algorithm through data assimilation techniqueDynamic error correction framework capable of improving the streamflow forecast at longer lead time is proposedOverall the developed framework improves the forecast accuracy along with quantifying the model prediction uncertainty [ABSTRACT FROM AUTHOR]

This paper compares the results of multiaxial fatigue life estimation using machine learning methods and classical fatigue models. The fatigue life of PA38‐T6 aluminum alloy under uniaxial, proportional, and non‐proportional loading, including asynchronous loading, is studied. Machine learning methods are trained only on basic loadings, namely, axial, torsional, and 90° out‐of‐phase. The results obtained with the machine learning algorithms, dense neural networks, support vector regression (with linear and radial basis functions), decision tree, random forest, and XGBoost algorithms, are comparable to the results of classical models like Ellyin–Gołoś, Fatemi–Socie, and Ince–Glinka. The best results are achieved for dense neural networks. In that case, they are often slightly more accurate than those obtained using classical methods. Fatigue life prediction based on machine learning and fatigue models were compared.Solely axial, torsional, and out‐of‐phase data were used to learn algorithms.Machine learning‐based results are comparable to those achieved by fatigue models.The multiaxial fatigue life was predicted most accurately by dense neural networks. [ABSTRACT FROM AUTHOR]

Summary: Machine learning process is an important part in intrusion detection system (IDS) and plays an essential role in security information. Analysts operate IDS for the study of server logs or packets of data to identify network traffic malware. IDS automates this process with machine learning‐based techniques that allow packet detection on a continuous basis without much human effort or intervention through network analyzing data. It has the following limitations on low rate of detection, false alarm rate is high, and so on, to which the conventional IDS has indicated. We proposed a stacking model that is more accurate to detect the attacks than other techniques of Random Forest, Naïve Bayes, and Decision Tree C4.5. Different researchers are exploring various techniques to classify the rules for associations. But the measure of confidence itself, as investigated in various literature, produces many times inaccurate results. So based on a recent value test called the significant rule power factor (SRPF), this paper discusses classifier. The experiments are carried out by implementation in Weka's method. Results indicate that the proposed classifier is a newly created, light weight classifier with better accuracy performance. Detection of different attacks uses NSL‐KDD dataset to train and test the data in this technique. This proposed work is measured using different performance metrics to other intrusion detection models. Compared to existing methods, our proposed model offers prominent increase in detection accuracy. [ABSTRACT FROM AUTHOR]

Class imbalance is a common and critical challenge in machine learning classification problems, resulting in low prediction accuracy. While numerous methods, especially data augmentation methods, have been proposed to address this issue, a method that works well on one dataset may perform poorly on another. To the best of our knowledge, there is still no one single best approach for handling class imbalance that can be uniformly applied. In this paper, we propose an approach named smart data augmentation (SDA), which aims to augment imbalanced data in an optimal way to maximize downstream classification accuracy. The key novelty of SDA is an equation that can bring about an augmentation method that provides a unified representation of existing sampling methods for handling multi‐level class imbalance and allows easy fine‐tuning. This framework allows SDA to be seen as a generalization of traditional methods, which in turn can be viewed as specific cases of SDA. Empirical results on a wide range of datasets demonstrate that SDA could significantly improve the performance of the most popular classifiers such as random forest, multi‐layer perceptron, and histogram‐based gradient boosting. [ABSTRACT FROM AUTHOR]

Multiple tools exist for separately simulating and estimating the parameters of system dynamics models. Artificial intelligence (AI) has been increasingly used to estimate the parameters of system dynamics models. The development of modeling tools and advanced environments has resulted in great benefits to the community at large. The incorporation of AI tools into system dynamics presents opportunities for expanding on current decision‐making methods. As systems become complex, the need to incorporate evidence‐based data‐driven methods increases. By integrating system dynamics tools and facilitating AI and system dynamics simulation in an integrated environment, model parameters can be estimated with the latest data, and the integrity of the model can be retained effectively. This provides an advantage to the efficiency and capabilities of the system dynamics model and its analysis. This paper presents a general methodology to incorporate regression AI into system dynamics models for simulation and analysis. To demonstrate the validity of the methodology, a case study involving a susceptible‐infected‐recovered model and empirical data from the COVID‐19 pandemic is performed using support vector machines (SVMs), artificial neural networks (ANNs), and random forests. [ABSTRACT FROM AUTHOR]

Very high‐cycle fatigue life (VHCF) prediction of high‐strength steel based on machine learning (ML) was investigated in this paper. First, a total of 173 sets of experimental data on the VHCF of high‐strength steel were collected to train the ML model. The sensitivity coefficient analysis indicated that inclusion size and maximum stress were the strongest correlation parameters with fatigue life and selected as the input features for the final model training. The S–N curve predicted by the obtained ML model closely aligns with the actual S–N curve. Among the three algorithm models, namely, random forest, XG boost, and gradient boosting, the gradient boosting model exhibited superior performance and achieved the highest accuracy in predicting the VHCF life of high‐strength steel. A comparison between the Murakami model and the gradient boosting model was conducted. It is indicated that ML exhibits superior predictive performance with high efficiency and excellent accuracy. Highlights: S–N curve predicted by machine learning model closely aligns with that by experiments.Gradient boosting model exhibited superior performance in predicting the VHCF life.Machine learning model outperforms the Murakami model in the terms of accuracy. [ABSTRACT FROM AUTHOR]

Wireless communication devices must be protected from malicious threats, including active jamming attacks, due to the widespread use of wireless systems throughout our every‐day lives. Jamming mitigation techniques are predominately evaluated through simulation or with hardware for very specific jamming conditions. In this paper, an experimental software defined radio‐based RF jamming mitigation platform which performs online jammer classification and leverages reconfigurable beam‐steering antennas at the physical layer is introduced. A ray‐tracing emulation system is presented and validated to enable hardware‐in‐the‐loop jamming experiments of complex outdoor and mobile site‐specific scenarios. Random forests classifiers are trained based on over‐the‐air collected data and integrated into the platform. The mitigation system is evaluated for both over‐the‐air and ray‐tracing emulated environments. The experimental results highlight the benefit of using the jamming mitigation system in the presence of active jamming attacks. [ABSTRACT FROM AUTHOR]

In recent years, early detection of hepatitis C virus (HCV) disease has been a vital task in the medical science field. HCV became the main health concern to the public, as it was noticed to have more blood donors in Egypt equated to other nationalities. The WHO assessed that in 2019, around 290 000 individuals died from hepatitis C, which says the seriousness of the HCV disease. So, early prediction, preventions, and curing the disease are vital components to save individuals from HCV. In this paper, we propose and investigate experimental results of the five machine learning (ML) models and probabilistic neural network (PNN) based approach to detect the HCV utilizing University of Califonia Irvine (UCI) ML Egyptian HCV dataset. We also analyze the statistical reports of HCV stages and their features. As per result analysis, the random forest (RF) ML model performs superiorly to other traditional ML algorithms with 97.5% of accuracy. The PNN (incremental hidden layer [HL] neurons) based proposal model shows a very high performance (99.6% of accuracy) at 30‐HL neurons of PNN. As per comparative analysis, the proposed model is superior to experimental basic ML models, and early HCV disease works are related to this area. This research focuses on early detection, prevention, and challenges of handling HCV. [ABSTRACT FROM AUTHOR]

Real estate price prediction takes an important role in the economy that can drive up and down the stock prices and even generate disruptive economic events. Many researchers have tried to understand the pricing mechanism with machine learning techniques such as support vector machine, neural network, random forest, and AdaBoost. The boundary problem, on the other hand, makes the pricing scheme more complicated, and this trend is accelerated especially in the situation of population decline in Japan. In this paper, we discuss how we could approach the boundary problem in real estate prediction. We propose a new comprehensive inference model extending and adapting XGBoost to the domain that has the boundary conditions problem by utilizing the distance between the instances in the domain data set to make the layers of bumpy boundaries smooth for more accurate predictions and robustness against the domain data set. The experiments result showed our proposed method performed well on both hypothetical data sets and actual real estate price data. © 2022 Institute of Electrical Engineers of Japan. Published by Wiley Periodicals LLC. [ABSTRACT FROM AUTHOR]

Chronic kidney disease is a global health issue that affects millions of people worldwide and causes significant social, economic, and medical issues. Several automated detection systems can diagnose chronic kidney disease. This paper proposes the recursive random forest feature selection (RFFS) based ensemble learning algorithm to diagnose chronic kidney diseases (CKD). In the decision point, decision tree‐based classifiers are used. The accuracy and kappa scores are used to determine the classification results. According to the results of the proposed algorithm's performance analyses, the ensemble learning classifiers outperform other classifiers for classifying CKD. The proposed RFFS algorithm achieves 6%–40% improved prediction accuracy using the feature selection algorithm. Further, it attains 15%–39% of reduced mean square error. The performance metrics, precision, sensitivity, specificity, f1 score, and Jaccard scores, have been analysed and show greater results for the RFFS algorithm. Thus, the proposed RFE‐GB algorithm results prove it as a prominent model for CKD estimation and treatment. [ABSTRACT FROM AUTHOR]

Low‐cost air quality monitoring networks can potentially increase the availability of high‐resolution monitoring to inform analytic and evidence‐informed approaches to better manage air quality. This is particularly relevant in low and middle‐income settings where access to traditional reference‐grade monitoring networks remains a challenge. However, low‐cost air quality sensors are impacted by ambient conditions which could lead to over‐ or underestimation of pollution concentrations and thus require field calibration to improve their accuracy and reliability. In this paper, we demonstrate the feasibility of using machine learning methods for large‐scale calibration of AirQo sensors, low‐cost PM sensors custom‐designed for and deployed in Sub‐Saharan urban settings. The performance of various machine learning methods is assessed by comparing model corrected PM using k‐nearest neighbours, support vector regression, multivariate linear regression, ridge regression, lasso regression, elastic net regression, XGBoost, multilayer perceptron, random forest and gradient boosting with collocated reference PM concentrations from a Beta Attenuation Monitor (BAM). To this end, random forest and lasso regression models were superior for PM2.5 and PM10 calibration, respectively. Employing the random forest model decreased RMSE of raw data from 18.6 μg/m3 to 7.2 μg/m3 with an average BAM PM2.5 concentration of 37.8 μg/m3 while the lasso regression model decreased RMSE from 13.4 μg/m3 to 7.9 μg/m3 with an average BAM PM10 concentration of 51.1 μg/m3. We validate our models through cross‐unit and cross‐site validation, allowing analysis of AirQo devices' consistency. The resulting calibration models were deployed to the entire large‐scale air quality monitoring network consisting of over 120 AirQo devices, which demonstrates the use of machine learning systems to address practical challenges in a developing world setting. [ABSTRACT FROM AUTHOR]

In this paper we address the impact of the introduction of the National Greenhouse and Energy Reporting scheme on corporate carbon reporting, and subsequently identify factors that influence the level of voluntary carbon reporting. A review of the literature demonstrates a large number of potential factors have been previously deployed to explain voluntary reporting practices; however, the analytical and empirical methods widely used in the literature have limiting statistical assumptions and confine analysis to a small number of explanatory factors. To address this limitation in prior research we apply advanced machine learning methods, such as gradient boosting machines and random forests, to identify predictive variables through analytical means. We compare the performance of machine learning methods with traditional methods such as logistic regression. We find that machine learning methods significantly outperform logistic regression and provide fundamentally different interpretations of the role and influence of different predictive variables on voluntary carbon reporting. While most variables were not statistically significant in the logit results, a number of key proxies for financial performance, corporate governance, and corporate social responsibility have out‐of‐sample predictive power of the level of voluntary carbon reporting in the machine learning analysis. [ABSTRACT FROM AUTHOR]