Search

Regression testing is required in each iteration of microservice systems. Regression testing selection, which reduces testing costs by selecting a subset from the original test cases, is one of the main techniques to optimize regression testing. Existing techniques mainly rely on the information retrieved from artifacts such as code files and system models. For microservice systems with service autonomy, development method diversity and a large amount of services, such artifacts are too difficultly obtained and costly processed to apply those approaches. This paper presents a regression testing selection approach called MRTS-BP, which needs the API gateway layer logs instead of code files and system models as inputs. By parsing the API gateway layer logs, our approach establishes the service dependency matrix, which in further is transformed into a directed graph with the services as nodes. Then, to find out which test cases are affected by service changes, an algorithm based on belief propagation is presented to compute the quantitative results of service-change propagation from the directed graph. Finally, the relationships between original test cases and service-change propagation results are established to select test cases with three strategies. To evaluate the efficiency of MRTS-BP, the empirical study based on four microservice systems is presented. A typical technique RTS-CFG is compared with MRTS-CFG and four experiments are setup to investigate four research questions. The results show that MRTS-BP can not only reduce the number of test cases by half compared with the retest-all strategy while ensuring the safety, but also save at least 20% testing time costs more than that of RTS-CFG. MRTS-BP is more practical than the techniques relying on the artifacts when the latter cannot be implemented due to the artifacts are difficult to obtain and process.

Deep learning models have been very successful in the application of machine learning methods, often out-performing classical statistical models such as linear regression models or generalized linear models. On the other hand, deep learning models are often criticized for not being explainable nor allowing for variable selection. There are two different ways of dealing with this problem, either we use post-hoc model interpretability methods or we design specific deep learning architectures that allow for (more) easy interpretation and explanation. This paper builds on our previous work on the LocalGLMnet architecture that gives an interpretable deep learning architecture. In the present paper, we show how group LASSO regularization (and other regularization schemes) can be implemented within the LocalGLMnet architecture so that we receive feature sparsity for variable selection. We benchmark our approach with the recently developed LassoNet of Lemhadri et al.

Remote user authentication schemes provide a system to verify the legitimacy of remote users’ authentication request over insecure communication channel. In last years, many authentication schemes using password and smart card have been proposed. However, password might be revealed or forgotten and smart card might be shared, lost or stolen. In contrast, the biometrics, such as face, fingerprint or iris, have no such weakness. With the trend of mobile payment, more and more applications of mobile payment use biometrics to replace password and smart card. In this paper, we propose a biometric-based remote authentication scheme substituting biometric and mobile device bounded by user for password and smart card. This scheme is more convenient, suitable and securer than the schemes using smart cards on mobile payment environment.

Stroke is a leading cause of neurological injury characterized by impairments in multiple neurological domains including cognition, language, sensory and motor functions. Clinical recovery in these domains is tracked using a wide range of measures that may be continuous, ordinal, interval or categorical in nature, which presents challenges for standard multivariate regression approaches. This has hindered stroke researchers' ability to achieve an integrated picture of the complex time-evolving interactions amongst symptoms. Here we use tools from network science and machine learning that are particularly well-suited to extracting underlying patterns in such data, and may assist in prediction of recovery patterns. To demonstrate the utility of this approach, we analyzed data from the NINDS tPA trial using the Trajectory Profile Clustering (TPC) method to identify distinct stroke recovery patterns for 11 different neurological domains at 5 discrete time points. Our analysis identified 3 distinct stroke trajectory profiles that align with clinically relevant stroke syndromes, characterized both by distinct clusters of symptoms, as well as differing degrees of symptom severity. We then validated our approach using graph neural networks to determine how well our model performed predictively for stratifying patients into these trajectory profiles at early vs. later time points post-stroke. We demonstrate that trajectory profile clustering is an effective method for identifying clinically relevant recovery subtypes in multidimensional longitudinal datasets, and for early prediction of symptom progression subtypes in individual patients. This paper is the first work introducing network trajectory approaches for stroke recovery phenotyping, and is aimed at enhancing the translation of such novel computational approaches for practical clinical application., Comment: 20 pages, 5 figures

Machine Learning (ML) solutions are nowadays distributed, according to the so-called server/worker architecture. One server holds the model parameters while several workers train the model. Clearly, such architecture is prone to various types of component failures, which can be all encompassed within the spectrum of a Byzantine behavior. Several approaches have been proposed recently to tolerate Byzantine workers. Yet all require trusting a central parameter server. We initiate in this paper the study of the ``general'' Byzantine-resilient distributed machine learning problem where no individual component is trusted. We show that this problem can be solved in an asynchronous system, despite the presence of $\frac{1}{3}$ Byzantine parameter servers and $\frac{1}{3}$ Byzantine workers (which is optimal). We present a new algorithm, ByzSGD, which solves the general Byzantine-resilient distributed machine learning problem by relying on three major schemes. The first, Scatter/Gather, is a communication scheme whose goal is to bound the maximum drift among models on correct servers. The second, Distributed Median Contraction (DMC), leverages the geometric properties of the median in high dimensional spaces to bring parameters within the correct servers back close to each other, ensuring learning convergence. The third, Minimum-Diameter Averaging (MDA), is a statistically-robust gradient aggregation rule whose goal is to tolerate Byzantine workers. MDA requires loose bound on the variance of non-Byzantine gradient estimates, compared to existing alternatives (e.g., Krum). Interestingly, ByzSGD ensures Byzantine resilience without adding communication rounds (on a normal path), compared to vanilla non-Byzantine alternatives. ByzSGD requires, however, a larger number of messages which, we show, can be reduced if we assume synchrony., This is a merge of arXiv:1905.03853 and arXiv:1911.07537; arXiv:1911.07537 will be retracted

Gait is an individual biometric behavior which can be detected based on distance which has different submissions in social security, forensic detection and crime prevention. Hence, in this paper, Advanced Deep Belief Neural Network with Black Widow Optimization (ADBNN-BWO) Algorithm is developed to identify the human emotions by human walking style images. This proposed methodology is working based on four stages like pre-processing, feature extraction, feature selection and classification. For the pre-processing, contrast enhancement median filter is used and Hu Moments, GLCM, Fast Scale-invariant feature transform (F-SIFT), in addition skeleton features are used for the feature extraction. To extract the features efficiently, the feature extraction algorithm can be often very essential calculation. After that, feature selection is performed. Then the classification process is done by utilizing the proposed ADBNN-BWO Algorithm. Based on the proposed method, the human gait recognition is achieved which utilized to identify the emotions from the walking style. The proposed method is validated by using the open source gait databases. The proposed method is implemented in MATLAB platform and their corresponding performances/outputs are evaluated. Moreover, the statistical measures of proposed method are also determined and compared with the existing method as Artificial Neural Network (ANN), Mayfly algorithm with Particle Swarm Optimization (MA-PSO), Recurrent Neural Network -PSO (RNN-PSO) and Adaptive Neuro Fuzzy Inference System (ANFIS) respectively.

Fake news detection is a challenging problem in online social media, with considerable social and political impacts. Several methods have already been proposed for the automatic detection of fake news, which are often based on the statistical features of the content or context of news. In this paper, we propose a novel fake news detection method based on Natural Language Inference (NLI) approach. Instead of using only statistical features of the content or context of the news, the proposed method exploits a human-like approach, which is based on inferring veracity using a set of reliable news. In this method, the related and similar news published in reputable news sources are used as auxiliary knowledge to infer the veracity of a given news item. We also collect and publish the first inference-based fake news detection dataset, called FNID, in two formats: the two-class version (FNID-FakeNewsNet) and the six-class version (FNID-LIAR). We use the NLI approach to boost several classical and deep machine learning models including Decision Tree, Naïve Bayes, Random Forest, Logistic Regression, k-Nearest Neighbors, Support Vector Machine, BiGRU, and BiLSTM along with different word embedding methods including Word2vec, GloVe, fastText, and BERT. The experiments show that the proposed method achieves 85.58% and 41.31% accuracies in the FNID-FakeNewsNet and FNID-LIAR datasets, respectively, which are 10.44% and 13.19% respective absolute improvements.

Due to Big Data and the Internet of Things, Machine Learning algorithms targeted specifically to model evolving data streams had gained attention from both academia and industry. Many Incremental Learning models had been successful in doing so, but most of them have one thing in common: they are complex variants of batch learning algorithms, which is a problem since, in a streaming setting, less complexity and more performance is desired. This paper proposes the Incremental LSTM model, which is a variant of the original LSTM with minor changes, that can tackle evolving data streams problems such as concept drift and the elasticity-plasticity dilemma without neither needing a dedicated drift detector nor a memory management system. It obtained great results that show it reacts fast to concept drifts and that is also robust to noise data.

Most current part libraries are created and deployed in specific usage environments or CAD platforms, which inevitably brings obstacles in the share and exchange for part information. To reduce repeated development and provide a uniform interface for designers in different sites, a 3D part library in cloud scheme is presented in this paper. Host programs with predetermined specification can access the part data through an adapter according to their customized requests with uniform interfaces, which constructs a ubiquitous service. To realize host independence, the part models are created in a native ACIS modeler, and then they are converted into 3D files in various formats for practical needs, finally these files are imported into CAD systems or other platforms in real designs. The whole framework can be divided into three components, namely, PLS(Part Library Service) provider, PLS adapters and hosts. PLS provider is the kernel of 3D data access service in cloud scheme, while PLS adapters serve as the bridges that connect PLS provider and hosts, and the PLS can be grafted on various applications including current mainstream CAD systems as a plug-in module or run on the websites or even mobile terminals. The PLS provider is deployed and maintained on cloud and users can acquire remote part information within a local ongoing project. In the detailed construction of this part library, diversiform knowledge for part parameters and structures is implanted to define the geometry and rule constraints in the 3D modeling, with which the backstage has the ability of conveniently editing the information in the part library for better upgrade and contrapuntally services. The concept has been implemented within a PaaS framework to provide the ubiquitous 3D part data access, which has been successfully applied in a large number of manufacturing enterprises, and accumulates considerable practical cases.

For a general production network, state-of-the-art methods for constructing sparse flexible designs are heuristic in nature, typically computing a proxy for the quality of unseen networks and using that estimate in a greedy manner to modify a current design. This paper develops two machine learning-based approaches to constructing sparse flexible designs that leverage a neural network to accurately and quickly predict the performance of large numbers of candidate designs. We demonstrate that our heuristics are competitive with existing approaches and produce high-quality solutions for both balanced and unbalanced networks. Finally, we introduce a novel application of process flexibility in healthcare operations to demonstrate the effectiveness of our approach in a large numerical case study. We study the flexibility of linear accelerators that deliver radiation to treat various types of cancer. We demonstrate how clinical constraints can be easily absorbed into the machine learning subroutine and how our sparse flexible treatment networks meet or beat the performance of those designed by state-of-the-art methods.

ImageNet has millions of images that are labeled with English WordNet synsets. This paper investigates the extension of ImageNet to Arabic using Arabic WordNet. The objective is to discover if Arabic synsets can be found for synsets used in ImageNet. The primary finding is the identification of Arabic synsets for 1,219 of the 21,841 synsets used in ImageNet, which represents 1.1 million images. By leveraging the parent-child structure of synsets in ImageNet, this dataset is extended to 10,462 synsets (and 7.1 million images) that have an Arabic label, which is either a match or a direct hypernym, and to 17,438 synsets (and 11 million images) when a hypernym of a hypernym is included. When all hypernyms for a node are considered, an Arabic synset is found for all but four synsets. This represents the major contribution of this work: a dataset of images that have Arabic labels for 99.9% of the images in ImageNet.

Nowcasts of strong convective precipitation and radar-based quantitative precipitation estimations have always been hot yet challenging issues in meteorological sciences. Data-driven machine learning, especially deep learning, provides a new technical approach for the quantitative estimation and forecasting of precipitation. A high-quality, large-sample, and labeled training dataset is critical for the successful application of machine-learning technology to a specific field. The present study develops a benchmark dataset that can be applied to machine learning for minute-scale quantitative precipitation estimation and forecasting (QpefBD), containing 231,978 samples of 3,185 heavy precipitation events that occurred from April to October during the years 2016 to 2018 in six provinces in central and eastern China. Each individual sample consists of eight products of weather radars at 6-min intervals within the time window of the corresponding event and products of 27 physical quantities at hourly intervals that describe the atmospheric dynamic and thermodynamic conditions. Two data labels, i.e., ground precipitation intensity and areal coverage of heavy precipitation at 6-minute intervals, are also included. The present study describes the basic components of the dataset and data processing and provides metrics for the evaluation of model performance on precipitation estimation and forecasting. Based on these evaluation metrics, some simple, commonly used methods are applied to evaluate precipitation estimates and forecasts. The results can serve as the benchmark reference for the performance evaluation of machine learning models using this dataset. This paper also gives some suggestions and scenarios of the QpefBD application. We believe that the application of this benchmark dataset will promote interdisciplinary collaboration between meteorological sciences and artificial intelligence sciences, providing a new way for the identification and forecast of heavy precipitation.

Like traditional single label learning, multi-label learning is also faced with the problem of dimensional disaster.Feature selection is an effective technique for dimensionality reduction and learning efficiency improvement of high-dimensional data. In this paper, Logistic regression, manifold learning and sparse regularization were combined to construct a joint framework for multi-label feature selection (LMFS). Firstly, the sparsity of the eigenweight matrix is constrained by the $L_{2,1}$-norm. Secondly, the feature manifold and label manifold can constrain the feature weight matrix to make it fit the data information and label information better. An iterative updating algorithm is designed and the convergence of the algorithm is proved.Finally, the LMFS algorithm is compared with DRMFS, SCLS and other algorithms on eight classical multi-label data sets. The experimental results show the effectiveness of LMFS algorithm.

A number of the people who have recovered from the acute effects of COVID-19 are facing long term sequelae from the infection. As the COVID-19 pandemic is still evolving, so is knowledge of the long-term effects of the virus on patients who still experience symptoms. Clearly, news media play a crucial role in distributing information and this distribution of information can, in turn, influence the actions of the public. The purpose of this study was to describe the content of news coverage of COVID-19 long haul symptoms currently posted on the internet. This study utilized Google News, a news aggregator service, and included the first 100 English language pieces of news. Video content and news article content were coded in depth for information on COVID-19 long haul symptoms. A total of 41% of news reports mentioned the length of time that the COVID-19 related symptoms persist. The length of time was reported to range from 1 month to more than 1 year. The symptom most commonly mentioned was tiredness or fatigue (74%), followed by difficulty breathing or shortness of breath (62 cases; 62%), and difficulty thinking or concentrating (50 cases; 50%). Other symptoms were mentioned less frequently. There were no statistically significant differences in any of the content including having video, written news reports, or both video and written news reports by source of the news reports based on consumer, professional, or television or internet-based news (p = .14). More complete coverage by online news media of the long-term effects of COVID-19 enhances public awareness of the post-acute syndromes, augments health providers' awareness of the range of chronic COVID-19 effects and the possibility of a second infection, increases the probability of patients' seeking and obtaining the proper care for their symptoms, and contributes to preventive actions for enhancing public health.

Despite the rapid growth of online shopping and research interest in the relationship between online and in-store shopping, national-level modeling and investigation of the demand for online shopping with a prediction focus remain limited in the literature. This paper differs from prior work and leverages two recent releases of the U.S. National Household Travel Survey (NHTS) data for 2009 and 2017 to develop machine learning (ML) models, specifically gradient boosting machine (GBM), for predicting household-level online shopping purchases. The NHTS data allow for not only conducting nationwide investigation but also at the level of households, which is more appropriate than at the individual level given the connected consumption and shopping needs of members in a household. We follow a systematic procedure for model development including employing Recursive Feature Elimination algorithm to select input variables (features) in order to reduce the risk of model overfitting and increase model explainability. Among several ML models, GBM is found to yield the best prediction accuracy. Extensive post-modeling investigation is conducted in a comparative manner between 2009 and 2017, including quantifying the importance of each input variable in predicting online shopping demand, and characterizing value-dependent relationships between demand and the input variables. In doing so, two latest advances in machine learning techniques, namely Shapley value-based feature importance and Accumulated Local Effects plots, are adopted to overcome inherent drawbacks of the popular techniques in current ML modeling. The modeling and investigation are performed at the national level, with a number of findings obtained. The models developed and insights gained can be used for online shopping-related freight demand generation and may also be considered for evaluating the potential impact of relevant policies on online shopping demand.

Information and Communication Technologies, to a long extent, have a major influence on our social life, economy as well as on worldwide security. Holistically, computer networks embrace the Information Technology. Although the world is never free from people having malicious intents i.e. cyber criminals, network intruders etc. To counter this, Intrusion Detection System (IDS) plays a very significant role in identifying the network intrusions by performing various data analysis tasks. In order to develop robust IDS with accuracy in intrusion detection, various papers have been published over the years using different classification techniques of Data Mining (DM) and Machine Learning (ML) based hybrid approach. The present paper is an in-depth analysis of two focal aspects of Network Intrusion Detection System that includes various pre-processing methods in the form of dimensionality reduction and an assortment of classification techniques. This paper also includes comparative algorithmic analysis of DM and ML techniques, which applied to design an intelligent IDS. An experiment al comparative analysis has been carried out in support the verdicts of this work using ‘Python’ language on ‘kddcup99’ dataset as benchmark . Experimental analysis had been done in which we had found more impact on dimensionality reduction and MLP performed well in the true classification to establish secure network. The motive behind this effort is to detect different kinds of malware as early as possible with accuracy, to provide enhanced observant among various existing techniques that may help the fascinated researchers for future potential works.

The calls for actions to combat climate change have resulted in renewable energy adoption and, consequently, an emerging renewable energy industry. Of course, this has also precipitated the need for more renewable energy expert grounded in hybrid renewable energy system (HRES) software to support the growing industry. Hence, selecting particular software for teaching renewable energy modules for the training of students at a university level is essential. The appraisal and selection of HRES sizing and optimisation software have become crucial for instructors and learners being prepared for the industry. This paper's central contribution is to provide academics, tutors, and researchers in the field of renewable energy system sizing with an idea on how to proficiently and effectively deploy multi-criteria decision-making (MCDM) approaches in HRES sizing and optimisation software selection problems. A framework that combines fuzzy entropy method and fuzzy-VlseKriterijumska Optimizacija I Kompromisno Resenje (VIKOR) was used to rank HRES simulation software. The performance of the framework was compared with that of the complex proportional assessment (COPRAS) method. The article observed a difference between the fuzzy VIKOR and COPRAS solution for ranking the most preferred software; the most preferred option using the Fuzzy VIKOR is the BCHP screening tool, and HOMER for the COPRAS method. They ranked the least preferred software as RETScreen (VIKOR method) and ORCED (COPRAS method).

Deep convolutional neural networks (DCNNs) have dominated as the best performers on almost all computer vision tasks over the past several years. However, it remains a major challenge to deploy these powerful DCNNs in resource-limited environments, such as embedded devices and smartphones. To this end, 1-bit CNNs have emerged as a feasible solution as they are much more resource-efficient. Unfortunately, they often suffer from a significant performance drop compared to their full-precision counterparts. In this paper, we propose a novel Bayesian Optimized compact 1-bit CNNs (BONNs) model, which has the advantage of Bayesian learning, to improve the performance of 1-bit CNNs significantly. BONNs incorporate the prior distributions of full-precision kernels, features, and filters into a Bayesian framework to construct 1-bit CNNs in a comprehensive end-to-end manner. The proposed Bayesian learning algorithms are well-founded and used to optimize the network simultaneously in different kernels, features, and filters, which largely improves the compactness and capacity of 1-bit CNNs. We further introduce a new Bayesian learning-based pruning method for 1-bit CNNs, which significantly increases the model efficiency with very competitive performance. This enables our method to be used in a variety of practical scenarios. Extensive experiments on the ImageNet, CIFAR, and LFW datasets show that BONNs achieve the best in classification performance compared to a variety of state-of-the-art 1-bit CNN models. In particular, BONN achieves a strong generalization performance on the object detection task.

We report the results of a novel protocol for running online experiments using a combination of an online experimental platform in parallel with web-conferencing software in two formats—with and without subject webcams—to improve subjects’ attention and engagement. We compare the results between our online sessions with the offline (lab) sessions of the same experiment. We find that both online formats lead to comparable subject characteristics and performance as the offline (lab) experiment. However, the webcam-on protocol has less noisy data, and hence better statistical power, than the protocol without a webcam. The webcam-on protocol can detect reasonable effect sizes with a comparable sample size as in the offline (lab) protocol. Supplementary Information The online version contains supplementary material available at 10.1007/s40881-021-00112-w.

Artificial Intelligence (AI) has rapidly developed over the past several decades. Several related AI approaches, such as Machine Learning (ML), have been applied to research on landscape patterns and ecological processes. Our goal was to review the methods of AI, particularly ML, used in studies related to landscape ecology and the main topics addressed. We aimed to assess the trend in the number of ML papers and the methods used therein, and provide a synopsis and prospectus of current use and future applications of ML in landscape ecology. We conducted a systematic literature search and selected 125 papers for review. These were examined and scored according to multiple criteria regarding methods and topic. We applied quantitative statistical methods, including cluster analysis based on titles, abstracts, and keywords and a non-metric multidimensional scaling based on attributes assigned during the review. We used Random Forests machine learning to describe the differences between identified clusters in terms of the topics and methods they included. The most frequent method found was Random Forests, but it is noteworthy to mention the increasing popularity of tools related to Deep Learning. The topics cover both ecologically oriented issues and the landscape-human interface. There has been a rapid increase in ML and AI methods in landscape ecology research, with Deep Learning and complex multi-step pipeline AI methods emerging in the last several years. The rapid increase in the number of ML papers in landscape ecology research, and the range of methods employed in them, suggest explosive growth in application of these methods in landscape ecology. The increase of Deep Learning approaches in the most recent years suggest a major change in analytical paradigms and methodologies that we feel may transform the field and enable analyses of more complex pattern process relationships across vaster data sets than has been possible previously.

Nowadays, political parties have widely adopted social media for their party promotions and election campaigns. During the election, Twitter and other social media platforms are used for political coverage to promote the party and its candidates. This research discusses and estimates the stability of many volumetric social media approaches to forecast election results from social media activities. Numerous machine learning approaches are applied to opinions shared on social media for predicting election results. This paper presents a machine learning model based on sentiment analysis to predict Pakistan's general election results. In a general election, voters vote for their favorite party or candidate based on their personal interests. Social media has been extensively used for the campaign in Pakistan general election 2018. Using a machine learning technique, we provide a five-step process to analyze the overall election results, whether fair or unfair. The work is concluded with detailed experimental results and a discussion on the outcomes of sentiment analysis for real-world forecasting and approval for general elections in Pakistan.

Image encryption algorithms based on Chaotic approach are becoming increasingly popular for remotely sensed images using parallel techniques. It has been demonstrated that the most efficient image encryption algorithms are based on Chaos. Previous research using chaos-based cryptosystems has resulted in poor performance when using a single computer, compromising privacy, security, and reliability. Furthermore, there were issues when vulnerable satellite images were processed. This paper describes a novel chaos-based encryption technique that employs an external secret key and Henon, Logistic, and Gauss iterated maps. The proposed encryption algorithm is capable of efficiently encrypting a large number of images. When the number of images increases, however, these images become very small, and the technology becomes inefficient or impractical. This paper investigates the parallel method of image encryption on a large number of remotely sensed images in Hadoop. Hadoop's file visit method has been enhanced so that it can treat the entire Tiff file as a single unit. Furthermore, the file format is being extended to be supported by Hadoop in order to support GeoTiff in Hadoop. The results of the experiments show that the proposed parallel method for encryption is effective and scalable to a large number of images when compared to other well-known methods.

Fuzzy parameterized fuzzy soft matrices (fpfs-matrices) which can model problems involving fuzzy objects and parameters are one of the mathematical tools used to deal with decision-making problems. To utilize soft decision-making methods via fpfs-matrices in machine learning is likely to draw much scholarly attention. In this paper, we propose Comparison Matrix-Based Fuzzy Parameterized Fuzzy Soft Classifier (FPFS-CMC) in order to transfer modeling success of fpfs-matrices to machine learning. We then compare FPFS-CMC with Fuzzy Soft Set Classifier (FSSC), FussCyier, Fuzzy Soft Set Classification Using Hamming Distance (HDFSSC), and Fuzzy k-Nearest Neighbor (Fuzzy kNN) in consideration of accuracy, precision, recall, macro-F-score, and micro-F-score performance metrics, and 15 datasets in UCI Machine Learning Repository. Besides, we compare the proposed classifier with the state-of-the-art Support Vector Machine (SVM), Decision Tree (DT), and Adaptive Boosting (AdaBoost) in terms of five performance metrics herein. Afterward, the results from the experiments are analyzed by employing the Friedman and Nemenyi tests to assess the statistical significance of the differences in performances. Both experimental and statistical results show that FPFS-CMC outperforms the others. Finally, we provide the conclusive remarks and some suggestions for further research.

We have proposed MultiLexANFIS which is an adaptive neuro-fuzzy inference system (ANFIS) that incorporates inputs from multiple lexicons to perform sentiment analysis of social media posts. We classify tweets into two classes: neutral and non-neutral; the latter class includes both positive and negative polarity. This type of classification will be considered for applications that aim to test the neutrality of content posted by the users in social media platforms. In our proposed model, features are extracted by integrating natural language processing with fuzzy logic; hence, it is able to deal with the fuzziness of natural language in a very efficient and automatic manner. We have proposed a novel set of 64 rules for the proposed neuro-fuzzy network that can classify tweets correctly by working on fuzzy features fetched from VADER, AFINN and SentiWordNet lexicons. The proposed novel rules are domain independent, i.e., we can extend these rules for any textual data that employs lexicons. The antecedent and consequent parameters of the ANFIS are optimized by gradient descent and least squares estimate algorithms, respectively, in an iterative manner. The key contributions of this paper are: (1) a novel neuro-fuzzy system: MultiLexANFIS that takes as its input the positive and negative sentiment scores of tweets computed from multiple lexicons—VADER, AFINN and SentiWordNet, in order to classify the tweets into neutral and non-neutral content, (2) a novel set of 64 rules for the Sugeno-type fuzzy inference system—MultiLexANFIS, (3) single-lexicon-based ANFIS variants to classify tweets when multiple lexicons are not available and (4) comparison of MultiLexANFIS with different fuzzy, non-fuzzy and deep learning state of the art on various benchmark datasets revealing the superiority of our proposed neuro-fuzzy system for social sentiment analysis.

Software design patterns are well-known solutions for solving commonly occurring problems in software design. Detecting design patterns used in the code can help to understand the structure and behavior of the software, evaluate the quality of the software, and trace important design decisions. To develop and maintain a software system, we need sufficient knowledge of design decisions and software implementation processes. However, the acquisition of knowledge related to design patterns used in complex software systems is a challenging, time-consuming, and costly task. Therefore, using a suitable method to detect the design patterns used in the code reduces software development and maintenance costs. In this paper, we proposed a new method based on conceptual signatures to improve the accuracy of design pattern detection. So we used the conceptual signatures based on the purpose of patterns to detect the patterns’ instances that conform to the standard structure of patterns, and cover more instances of patterns’ variants and implementation versions of the patterns and improve the accuracy of pattern detection. The proposed method is a specific process in two main phases. In the first phase, the conceptual signature and detection formula for each pattern is determined manually. Then in the second phase, each pattern in the code is detected in a semi-automatic process using the conceptual signature and pattern detection formula. To implement the proposed method, we focused on GoF design patterns and their variants. We evaluated the accuracy of our proposed method on five open-source projects, namely, Junit v3.7, JHotDraw v5.1, QuickUML 2001, JRefactory v2.6.24, and MapperXML v1.9.7. Also, we performed our experiments on a set of source codes containing the instances of GoF design patterns’ variants for a comprehensive and fair evaluation. The evaluation results indicate that the proposed method has improved the accuracy of design pattern detection in the code.

This paper discusses the progress made so far and future challenges in integrating the analysis of multiple Non-Functional Properties (NFP) (such as performance, schedulability, reliability, availability, scalability, security, safety, and maintainability) into the Model-Driven Engineering (MDE) process. The goal is to guide the design choices from an early stage and to ensure that the system under construction will meet its non-functional requirements. The evaluation of the NFPs considered in this paper uses various kinds of NFP analysis models (also known as quality models) based on existent formalisms and tools developed over the years. Examples are queueing networks, stochastic Petri nets, stochastic process algebras, Markov chains, fault trees, probabilistic time automata, etc. In the MDE context, these models are automatically derived by model transformations from the software models built for development. Developing software systems that exhibit a good trade-off between multiple NFPs is difficult because the design of the software under construction and its underlying platforms have a large number of degrees of freedom spanning a very large discontinuous design space, which cannot be exhaustively explored. Another challenge in balancing the NFPs of a system under construction is due to the fact that some NFPs are conflicting—when one gets better the other gets worse—so an appropriate software process is needed to evaluate and balance all the non-functional requirements. The integration approach discussed in this paper is based on an ecosystem of inter-related heterogeneous modeling artifacts intended to support the following features: feedback of analysis results, consistent co-evolution of the software and analysis models, cross-model traceability, incremental propagation of changes across models, (semi)automated software process steps, and metaheuristics for reducing the design space size to be explored.

In recent times, the Internet of Things and big data analytics have become a hot research topic in mobile edge computing (MEC) desires wide-ranging research works for intelligent decision making. In this view, this paper designs an intelligent outlier detection with machine learning empowered big data analytics (IODML-BDA) for MEC. The proposed model involves an adaptive synthetic sampling-based outlier detection technique to eradicate the existence of outliers. Besides, the oppositional swallow swarm optimization (OSSO) based feature selection technique is used to choose an effective set of features. Finally, long short-term memory based classification model is employed to identify different class labels. The design of OSSO algorithm for feature selection with ADASYN technique for big data analytics show the novelty of the work. A comprehensive experimental analysis is carried out on GPS trajectories, movement prediction, water treatment plant, hepatitis, and Twitter datasets to confirm the experimental results. The experimentation outcomes pointed out that the proposed IODML-BDA model achieves the higher accuracy of 0.9735, 0.9816, 0.9798, 0.9896, and 0.9912 on the applied datasets.

Sentiment analysis is a type of contextual text mining that assesses how users feel about emotive topics that are frequently discussed on social media. To analyze the sentiments of the textual data, a number of sentiment analysis methods such as lexicon-based, machine learning-based, and hybrid methods have been proposed. Among all methods, unsupervised methods, especially clustering methods are generally preferred, as they can directly be applied over the unlabelled datasets. Therefore, in this paper, a roulette wheel-based cuckoo search clustering method has been proposed for sentiment analysis. The proposed clustering method finds the optimal cluster centroids from the contents of sentimental datasets which are further used for determining the sentiment polarity of a document. The efficiency of the proposed roulette wheel cuckoo search clustering method has been evaluated on nine sentimental datasets including Twitter and Spam review datasets and compared with K-means, cuckoo search, grey wolf optimizer, grey wolf optimizer with simulated annealing, hybrid step size based cuckoo search, and spiral cuckoo search. The experimental analysis shows that the proposed methods attain the best mean accuracy, mean precision, and mean recall over 80% of the datasets. To statistically validate the efficacy of the proposed approach, box plots and paired t-test are also carried out. From the statistical analysis and experimental findings, the efficacy of the proposed method can be observed. The proposed clustering approach has theoretical implications for further studies to examine the sentimental data. Furthermore, the proposed method has significant practical implications for establishing a system that can generate conclusive comments on any societal issue.

It is challenging for complex urban transportation networks to recommend taxi waiting spots for mobile passengers because the traditional centralized mining platform cannot address the storage and calculation problems of GPS trajectory big data, and especially the boundary identification of DBSCAN is difficult on the Spark parallel processing framework. To this end, we propose a parallel DBSCAN optimization algorithm with the silhouette coefficient and the pickup rate on Spark in this paper, named SP-DBSCAN, where users merely input one parameter to complete the distributed recommendation of the best waiting spot. Specifically, under the Hadoop distributed computing platform, a general framework of distributed modeling for waiting spot recommendation on Spark is developed to solve the distributed storage and parallel computing issues of the serial algorithm in handling data partition and clustering of large-scale traffic data on a single machine. Moreover, we put forward a parallel SP-DBSCAN algorithm on Spark to recommend the best waiting spot for passengers, where the traditional DBSCAN algorithm is optimized via the silhouette coefficient and the boarding ratio to address the parameter sensitive problem and the issue of the center of the non-convex clustering graph is solved by giving one cluster with two centroids in the clustering hotspot areas. Finally, experimental results on four groups of real-world taxi GPS trajectory data sets demonstrate that compared with C-DBSCAN and P-DBSCAN, the recognition rate of SP-DBSCAN is increased by 1.6%, 6.2%, 3.47%, and 5.8%, respectively. The empirical study indicates that the clustering region generated by our SP-DBSCAN algorithm can satisfy the requirements that passengers can ride in the hotspot area when they have not successfully hitchhiked at a specific location and turned to the next spot randomly.

This paper proposes a visual management scheme of medical things with a color-change radio frequency identification (RFID) tag. The color-change RFID tag employs a specific RFID tag integrated circuit (IC) and a laminated pH-indicating paper. The IC has energy harvesting and switched ground functions, which enable it to generate electricity to the laminated pH-indicating paper. This phenomenon causes electrolysis of NaCl solution absorbed in the laminated pH-indicating paper. Electrolysis generates alkaline matter to change the color of the pH-indicating paper. This paper gives a new and sensitive structure of the laminated pH-indicating paper. The proposed advanced color-change RFID tag with new laminated pH-indicating paper succeeds in changing its color noticeably at a 1 m distance using an RFID reader radiating 1 W radio waves. The color change was observed 3–5 s after starting radio wave irradiation. The results of this experiment also confirm that the changed color can be held for over 24 h. Furthermore, two demonstrations of the visual management system of medical things (patient clothes and sanitizers) are presented.

Cell studies play an important role in the basis of studies on cancer diagnosis and treatment. Reliable viability assays on cancer cell studies are essential for the development of effective drugs. Lens-free digital in-line holographic microscopy has become a powerful tool in the characterization and viability analysis of microparticles such as cancer cells due to its advantages such as high efficiency, low cost, and flexibility to integrate with other components. This study is designed to perform viability tests using fractal dimensions of alive and dead cancer cells based on digital holographic microscopy and machine learning. In the in-line holography configuration, a microscopy assembly consisting of inexpensive components was built using an LED source, and the images were reconstructed using computational methods. The standard US Air Force Resolution Target was used to evaluate the capability of our imaging setup then holograms of stained cancer cells were recorded. To characterize individual cells, 19 different rotational invariant fractal dimension values were extracted from the images as features. An artificial neural network technique was employed for the classification of fractal features extracted from cells. The artificial neural network was compared with four other machine learning techniques through five different classification performance measures. The empirical results indicated that artificial neural networks performed better than compared classification techniques with accuracies of 99.65%. The method proposed in this paper provides a new method for the study of cell viability which has the advantages of high accuracy and potential for laboratory application.

In recent days, data-driven corporate landscape data visualization has been essential and regularly used to help make decisions closely connected to many e-commerce firms' main profits. Since there is an increased need for database professionals to support reliable and successful data visualization, the high demand for data processing is based on data volume, speed, and data integrity. Currently, large-data visualization applications are complicated and varied. There are no standardized and uniform application creation methods, resulting in low visualization applications becoming reusable and challenging to extend and sustain. This paper suggests an Artificial Neural Network (ANN) based High Dimensional Data Visualization Technique for Interactive Data Exploration (HD2VT-IDE) in e-commerce. In this study, the ANN applies to the e-commerce product advertising recommendation for predicting the advertising click rate. The proposed framework has been evaluated with the existing models and observed the highest performance in terms of prediction accuracy (62.4%) and customer consistency rate (92.1%).

Establishing reliable correspondences plays a vital role in many feature-matching based computer vision tasks. Given putative correspondences of feature points in two images, in this paper, we propose a novel network for inferring the probabilities of correspondences being inliers or outliers and regressing the relative pose encoded by the essential matrix. Previous research proposed an end-to-end permutation-equivariant classification network based on multi-layer perceptrons and context normalization. However, the context normalization treats each correspondence equally and ignore the extraction of channel information, as a result the representation capability of potential inliers can be reduced. To solve this problem, we apply attention mechanism in our network to capture complex information of the feature maps. Specifically, we introduce two types of attention blocks. We adopt the spatial attention block to capture complex spatial contextual information, and the rich channel information can be obtained by utilizing the channel attention block. To obtain richer contextual information and feature maps with stronger representative capacity, We combine these attention blocks with the PointCN block to form a new network with strong representative ability. Experimental results on several benchmark datasets show that the performance on outlier removal and camera pose estimation is significantly improved over the state-of-the-arts.

As social media has become a ubiquitous part of daily life, researchers made a great progress in identifying the emotion in user-generated texts. However, it is a challenging task as people express their emotion in explicit and implicit ways. This paper focuses on the problem of identifying sentiments from implicit sentences which contain no emotional word or phrase. Most of the existing sentiment classification models cannot identify the sentiments accurately since they usually focus on extracting features from grammatical information without taking contextual information into account. In this paper, we argue that the contextual information is the key to identify sentiments in implicit sentences. Moreover, multiple features extracting from different aspects should be taken into account to improve sentiment identification. This paper proposes a multi-feature neural network model considering three aspects: contextual information, syntactic information and semantic information. To better get the semantic information of the sentence, we propose an attention mechanism based on contextual affective space. The experimental results on the SMP2019-ECISA dataset demonstrate that our model outperforms the previous systems and strong neural baselines.

With the continuous progress and development of modern social science management and network security technology in China, the wide application of new-generation technologies such as intelligent robotics learning has made it more and more extensive in the field of analysis. At present, it has fully played an important technical leading role in many application fields, such as analysis methods, collection and sorting ability, analysis of various large-scale business computing application statistical data of enterprises, mining theoretical knowledge of natural environment science, image recognition, retrieval of global geographic location information and so on. Through the simulation of soil erosion in mountainous areas in my country, dynamic data analysis describes the basic situation of soil erosion in various mountainous areas in my country and the whole process of its evolution and development and comprehensively analyzes and studies various soil erosion events in mountainous areas in my country. The corresponding solutions can effectively and accurately respond to emergencies of soil erosion in mountainous areas and effectively reduce the loss of soil and water crops and economic losses in natural disasters. In recent years, information technology companies related to computer and machine learning capabilities, especially based on machine cloud computing, learning science and technology based on the theory of human machine artificial intelligence, and combining related professional technologies, have been used in various areas to prevent floods, droughts, risks, and rain in various areas. Disaster reduction and other non-natural engineering construction technology prevention measures have gradually been applied. The main research question of this paper is to design a related human body trajectory image detection system for new human facial image processing based on machine learning, and information collection and detection management system, and conduct in-depth research on its application prospects.

Artificial intelligence (AI) and Machine learning (ML) train machines to achieve a high level of cognition and perform human-like analysis. Both AI and ML seemingly fit into our daily lives as well as complex and interdisciplinary fields. With the rise of commercial, open-source, and user-catered AI/ML tools, a key question often arises whenever AI/ML is applied to explore a phenomenon or a scenario: what constitutes a good AI/ML model? Keeping in mind that a proper answer to this question depends on various factors, this work presumes that a goodmodel optimally performs and best describes the phenomenon on hand. From this perspective, identifying proper assessment metrics to evaluate the performance of AI/ML models is not only necessary but is also warranted. As such, this paper examines 78 of the most commonly-used performance fitness and error metrics for regression and classification algorithms, with emphasis on engineering and sciences applications.

A familiar argument goes as follows: natural languages have infinitely many sentences, finite representation of infinite sets requires recursion; therefore any adequate account of linguistic competence will require some kind of recursive device. The first part of this paper argues that this argument is not convincing. The second part argues that it was not the original reason recursive devices were introduced into generative linguistics. The real basis for the use of recursive devices stems from a deeper philosophical concern; a grammar that functions merely as a metalanguage would not be explanatorily adequate as it would merely push the problem of explaining linguistic competence back to another level. The paper traces this concern from Zellig Harris and Chomsky’s early work in generative linguistics and presents some implications.

SSL certificates hold immense importance when it comes to the security of the WebPKI. The trust in these certificates is driven by the strength of their cryptographic attributes and the presence of revocation features. In this paper, we perform a historical measurement study of cryptographic strength and the adoption of revocation mechanisms in the X.509 SSL certificates. In particular, it provides a real-world picture of the adoption of new certificate features and pushing new changes to the WebPKI ecosystem. We analyze the features like Online Certificate Status Protocol (OCSP) Stapling, RSA public key collisions, and the strength of certificate serial numbers. We observe the improvement in the adoption and reliability of these features for 2011–2020. Our analysis helps in identifying weaknesses and negligence in certificate issuance practices of Certificate Authorities such as lack of revocation, weak serial numbers, and issuance of the same public key across different certificates for different entities on the web known as the public key collision problem. Our results show that there is an overall increase of up to 97% in the adoption of OCSP-Stapling and OCSP extensions. Along with this, there are also significant improvements in the certificate serial number length with the top 6 CAs in our dataset issuing the majority of certificates with serial byte count greater than 30. We also discovered 803 public key collision sets in our dataset. To distinguish public key collisions, we provide a working criterion to distinguish permissible, safe collisions from unsafe, risky ones. Analysis of these features holds immense importance as weakness in any of these features could allow an adversary to forge certificate(s) and conduct several attacks examples of which include Flame malware, breach of the DigiNotar and Comodo certificate authorities.

Offline handwritten signature verification has been widely used for document forensics and biometrics, and it is a popular issue. Deep learning models have commonly been used to solve this problem. This research has two aims, and they are to present a high accuracy hybrid classification model for forensics and to collect and share a new handwritten signature dataset to contribute document forensics. In this paper, a novel deep signature verification model is presented. This method has three fundamental phases and they are deep feature generation using transfer learning, iterative minimum redundancy maximum relevance (IMRMR) feature selection, and classification phases. In the deep feature extraction phase, 13 pre-trained widely preferred convolutional neural networks (CNN) are selected. These are utilized as feature generators and 1000 features are extracted from each network. By merging the generated features, a feature vector with a length of 13,000 is created. This feature generation network is named Deep Feature Warehouse (DFW) since it uses 13 pre-trained deep feature extractors in the transfer learning model. The most valuable features of the DFW are selected by the proposed IMRMR method and the selected features are forwarded to the classifier. To test the proposed DFW and IMRMR based verification method, we collected a handwritten signature dataset and CEDAR dataset to obtain comparative results. The proposed DFW and ImRMR based document classification method reached 97.16 % classification accuracy on the collected dataset and 100 % accuracy on the CEDAR dataset. We have used two datasets to demonstrate the general classification ability of our proposal. The calculated results and findings obviously demonstrate the effectiveness of the proposed DFW and ImRMR image verification model. According to the results, our model has general success (it has developed on two datasets), and it is a lightweight machine learning model since it uses transfer learning for feature extraction.

Compared with other ecosystems, urban ecosystem is more vulnerable to the interference of external environmental factors, thus changing the original structure of urban ecosystem, so it is more vulnerable. In a city, how to reasonably layout the landscape of green space, to a certain extent, will affect the development of the city and people's daily life. The emergence of remote sensing technology provides convenience for the extraction of urban green space information. On this basis, scholars at home and abroad constantly put forward new theories and methods to improve the remote sensing technology. The accuracy of green space information extracted by high-resolution image is high, but the workload and cost of this method are large. The most important point is that some confidential information may be involved in the process of green space information extraction with high resolution, so this method is not applicable in the process of large-scale green space information extraction. At present, relevant data show that the medium resolution data can be used to extract urban green space information. In the planning of green space landscape in coastal cities, the city is divided into three areas: coastal water area, intermediate transition area, and land landscape area. According to the theory of landscape ecology, the city is divided into three parts: patch, matrix, and corridor. The process of urban ecological landscape planning is the process of communication between human and nature. In the city, people can feel the beauty of nature. According to the current relevant theory, there are many ways of plant configuration and configuration principles. This paper analyzes the methods and principles of plant configuration in the city by taking the road greening landscape design along a certain place as an example.

Deep learning-based video anomaly detection methods have drawn significant attention in the past few years due to their superior performance. However, almost all the leading methods for video anomaly detection rely on large-scale training datasets with long training times. As a result, many real-world video analysis tasks are still not applicable for fast deployment. On the other hand, the leading methods cannot provide interpretability due to the uninterpretable feature representations hiding the decision-making process when anomaly detection models are considered as a black box. However, the interpretability for anomaly detection is crucial since the corresponding response to the anomalies in the video is determined by their severity and nature. To tackle these problems, this paper proposes an efficient deep learning framework for video anomaly detection and provides explanations. The proposed framework uses pre-trained deep models to extract high-level concept and context features for training denoising autoencoder (DAE), requiring little training time (i.e., within 10 s on UCSD Pedestrian datasets) while achieving comparable detection performance to the leading methods. Furthermore, this framework presents the first video anomaly detection use of combing autoencoder and SHapley Additive exPlanations (SHAP) for model interpretability. The framework can explain each anomaly detection result in surveillance videos. In the experiments, we evaluate the proposed framework's effectiveness and efficiency while also explaining anomalies behind the autoencoder’s prediction. On the USCD Pedestrian datasets, the DAE achieved 85.9% AUC with a training time of 5 s on the USCD Ped1 and 92.4% AUC with a training time of 2.9 s on the UCSD Ped2.

Fault injection attacks (FIA), which cause information leakage by injecting intentional faults into the data or operations of devices, are one of the most powerful methods compromising the security of confidential data stored on these devices. Previous studies related to FIA report that attackers can skip instructions running on many devices through many means of fault injection. Most existing anti-FIA countermeasures on software are designed to secure against instruction skip (IS). On the other hand, recent studies report that attackers can use laser fault injection to manipulate instructions running on devices as they want. Although the previous studies have shown that instruction manipulation (IM) could attack the existing countermeasures against IS, no effective countermeasures against IM have been proposed. This paper is the first work tackling this problem, aiming to construct software-based countermeasures against IM faults. Evaluating program vulnerabilities to IM faults is required to consider countermeasures against IM faults. We propose three IM simulation environments for that aim and compare them to reveal their performance difference. GDB (GNU debugger)-based simulator that we newly propose in this paper outperforms the QEMU-based simulator that we presented in AICCSA:1–8, 2020 in advance, in terms of evaluation time at most $$\times$$ × 400 faster. Evaluating a target program using the proposed IM simulators reveals that the IM faults leading to attack successes are classified into four classes. We propose secure coding techniques as countermeasures against IMs of each four classes and show the effectiveness of the countermeasures using the IM simulators.

Estimating software enhancement efforts became a challenging task in software project management. Recent researches focused on identifying the best machine learning algorithms for software maintenance effort estimation. Most of the research publications investigated the use of ensemble learning for improving software effort estimation. Intending to increase the estimation accuracy over individual models, this paper investigates the use of the stacking ensemble method for estimating the enhancement maintenance effort (EME) of software projects. This paper makes a comparison between two machine learning-based approaches for estimating software EME: The M5P (as an individual model) and the stacking as an ensemble method combining different regression models (GBRegr, LinearSVR, and RFR) using the ISBSG dataset. A correlation-based feature selection (CFS) algorithm is basically used to achieve efficient data reduction. The selected ML techniques-based approaches were trained and tested on a dataset with relevant features leading to the improvement of estimate accuracy. Results show that the software EME estimation using CFS and stacking ensemble method is improved in terms of mean absolute error (MAE) = 0.0383 and root mean square error (RMSE) = 0.1973.

Data clustering is a crucial machine learning technique that helps divide a given dataset into many similar data objects where the data members resemble each other. It is an unsupervised learning algorithm and is hugely applied in different machine learning and data mining applications. k-means algorithm is one of the popular methods for clustering the data. However, this algorithm is not much suitable as it causes the problem of local entrapment. To resolve such issues, nature-inspired algorithms (NIAs) came into existence. Harris hawks optimizer (HHO) is a recently developed NIA inspired by the chasing and collaborative behavior of Harris hawks in real nature. The efficacy of HHO has already been proved by researchers in solving complex problems of different domains. In this paper, an opposition-based learning HHO (OHHO) is proposed for data clustering. The performance of OHHO is compared against six well-known algorithms on ten benchmark datasets of the UCI machine learning repository. Experimental values have justified the effectiveness of the proposed approach.

Time series data are a collection of chronological observations which are generated by several domains such as medical and financial fields. Over the years, different tasks such as classification, forecasting and clustering have been proposed to analyze this type of data. Time series data have been also used to study the effect of interventions overtime. Moreover, in many fields of science, learning the causal structure of dynamic systems and time series data is considered an interesting task which plays an important role in scientific discoveries. Estimating the effect of an intervention and identifying the causal relations from the data can be performed via causal inference. Existing surveys on time series discuss traditional tasks such as classification and forecasting or explain the details of the approaches proposed to solve a specific task. In this paper, we focus on two causal inference tasks, i.e., treatment effect estimation and causal discovery for time series data and provide a comprehensive review of the approaches in each task. Furthermore, we curate a list of commonly used evaluation metrics and datasets for each task and provide an in-depth insight. These metrics and datasets can serve as benchmark for research in the field.

In Bangladesh, groundwater is the main source of both drinking water and irrigation. Suction lift pumps and force mode of operation are the predominant technologies for groundwater abstraction in Bangladesh. For a sustainable usage policy, it is thus important to identify which technology would be more appropriate for which area in Bangladesh. With that aim in mind, this paper proposes a methodology that leverages the power of machine learning that can potentially learn intricate relationships between the (annual maximum) groundwater level (GWL) and the relevant hydrogeological factors (HGFs). A number of machine learning algorithms—both classification and regression models—was trained. Our classification models were trained as a binary classifier to predict the abstraction technology of a particular point. Notably, our best classification model was based on the Random Forest algorithm, which achieved an accuracy of 91% and an excellent value of 96% for the area under receiver operating characteristics curve, indicating its strong discriminant capability. We also identified (elevation derived from) digital elevation model, specific yield and lithology as the three most important HGFs for GWL in Bangladesh. On the other hand, to predict the actual (annual maximum) GWL, we employed a two-stage approach, where we first employed the above-mentioned classification model to identify the suitable abstraction technology for the point of interest and subsequently predict the actual GWL using the appropriate Random Forest regressor. This also had a reasonable accuracy (minimum absolute error was less than 1 for suction mode and less than 5 for the force mode). Finally, using our prediction models, we prepared groundwater (technology) maps for the whole Bangladesh.

International audience; Public Transportation Information Systems (PTIS) are widely used for public bus services amongst cities in the world. These systems gather information about trips, bus stops, bus speeds, ridership, etc. This massive data is an inviting source of information for machine learning predictive tools. However, it most often suffers from quality deficiencies, due to multiple data sets with multiple structures, to different infrastructures using incompatible technologies, to human errors or hardware failures. In this paper, we consider the impact of data cleansing on a classical machine-learning task: predicting urban bus commercial speed. We show that simple, transport specific business and quality rules can drastically enhance data quality, whereas more sophisticated rules may offer little improvements despite a high computational cost.; Les systèmes d'information pour les ransports publics (SITP) sont largement répandus et utilisés par les services de bus publics dans nobre de villes à travers le monde. Ces systèmes recueillent des informations sur les trajets, les arrêts de bus, la vitesse des bus, la fréquentation, etc. Ces données massives constituent une source d'information intéressante pour les outils prédictifs d'apprentissage automatique. Cependant, elles souffrent le plus souvent de déficiences qualitatives, dues à des ensembles de données multiples aux structures multiples, à des infrastructures différentes utilisant des technologies incompatibles, à des erreurs humaines ou à des défaillances matérielles. Dans cet article, nous examinons l'impact du nettoyage des données sur une tâche classique d'apprentissage automatique : la prédiction de la vitesse commerciale des bus urbains. Nous montrons que des règles de gestion et de qualité simples et spécifiques au transport peuvent améliorer considérablement la qualité des données, alors que des règles plus sophistiquées tendent à offrir peu d'améliorations tout en ayant un coût de calcul élevé.

With the rapid popularization of the Internet plus China and China’s rapid economic development, machine learning is playing an increasingly important role in model research. This paper presents a cold start method based on a hybrid Dirichlet process model and isolated forest. As we all know, the soil rock mixed slope is between the slope of soil and rock, most of its structure is loose and has large porosity, high permeability, and poor slope stability. Compared with soil slope and rock slope, the research on soil rock mixed slope is not comprehensive enough. Based on the study of geological environment conditions of highway slope, the mechanical properties of soil rock mixed slope are tested by traditional boundary equilibrium method and numerical simulation method. The stability of the three-dimensional slope is systematically studied in this paper. Nowadays, the mental health of left-behind children is still the focus of social attention. In this paper, a special school for left-behind children in a city is selected as the research object. The school is located between urban and rural areas. Since the children who stay in the school are from urban and rural areas, they will be selected as the research objects. Compared with ordinary children, left-behind children are more likely to have concerns, and their psychological changes are difficult to predict. What is more worrying is the long-term mental health problems. This paper aims to find the correlation between sports games and left-behind children’s mental health by introducing sports games in order to provide some methods and theoretical reference materials for solving the mental health problems of left-behind children.

Background With the rapid spread of COVID-19 worldwide, quick screening for possible COVID-19 patients has become the focus of international researchers. Recently, many deep learning-based Computed Tomography (CT) image/X-ray image fast screening models for potential COVID-19 patients have been proposed. However, the existing models still have two main problems. First, most of the existing supervised models are based on pre-trained model parameters. The pre-training model needs to be constructed on a dataset with features similar to those in COVID-19 X-ray images, which limits the construction and use of the model. Second, the number of categories based on the X-ray dataset of COVID-19 and other pneumonia patients is usually imbalanced. In addition, the quality is difficult to distinguish, leading to non-ideal results with the existing model in the multi-class classification COVID-19 recognition task. Moreover, no researchers have proposed a COVID-19 X-ray image learning model based on unsupervised meta-learning. Methods This paper first constructed an unsupervised meta-learning model for fast screening of COVID-19 patients (UMLF-COVID). This model does not require a pre-trained model, which solves the limitation problem of model construction, and the proposed unsupervised meta-learning framework solves the problem of sample imbalance and sample quality. Results The UMLF-COVID model is tested on two real datasets, each of which builds a three-category and four-category model. And the experimental results show that the accuracy of the UMLF-COVID model is 3–10% higher than that of the existing models. Conclusion In summary, we believe that the UMLF-COVID model is a good complement to COVID-19 X-ray fast screening models.

As the number of artificial intelligence (AI) applications increases rapidly and more people will be affected by AI’s decisions, there are real needs for novel AI systems that can deliver both accuracy and explanations. To address these needs, this paper proposes a new approach called eXplainable Mapping Analytical Process (XMAP). Different from existing works in explainable AI, XMAP is highly modularised and the interpretability for each step can be easily obtained and visualised. A number of core algorithms are developed in XMAP to capture the distributions and topological structures of data, define contexts that emerged from data, and build effective representations for classification tasks. The experiments show that XMAP can provide useful and interpretable insights across analytical steps. For the binary classification task, its predictive performance is very competitive as compared to advanced machine learning algorithms in the literature. In some large datasets, XMAP can even outperform black-box algorithms without losing its interpretability.