Search

This work presents a novel multivariate statistical algorithm, Decision Tree-PLS (DT-PLS), to improve the prediction and understanding of dynamic processes based on local partial least square regression (PLSR) models for characteristic process groups defined based on Decision Tree (DT) analysis. The DT-PLS algorithm is successfully applied to two different cell culture data sets, one obtained from bioreactors of 3.5 L lab scale and the other obtained from the 15 ml ambr microbioreactor system. Substantial improvement in the predictive capabilities of the model can be achieved based on the localization compared to the classical PLSR approach, which is implemented in the commercially available packages. Additionally, the differences in the model parameters of the local models suggest that the governing process variables vary for the different process regimes indicating the different states of the cell under different process conditions., (© 2019 American Institute of Chemical Engineers.)

Rationale, Aims, and Objectives: Randomization ensures that treatment groups do not differ systematically in their characteristics, thereby reducing threats to validity that may otherwise explain differences in outcomes. Large observed imbalances in patient characteristics may indicate that selection bias is being introduced into the treatment allocation process. We introduce classification tree analysis (CTA) as a novel algorithmic approach for identifying potential imbalances in characteristics and their interactions when provisionally assigning each new participant to one or the other treatment group. The participant is then permanently assigned to the treatment group that elicits either no or less imbalance than if assigned to the alternate group., Method: Using data on participant characteristics from a clinical trial, we compare 3 different treatment allocation approaches: permuted block randomization (the original allocation method), minimization, and CTA. Treatment allocation performance is assessed by examining balance of all 17 patient characteristics between study groups for each of the allocation techniques., Results: While all 3 treatment allocation techniques achieved excellent balance on main effect variables, Classification tree analysis further identified imbalances on interactions and in the distributions of some of the continuous variables., Conclusions: Classification tree analysis offers an algorithmic procedure that may be used with any randomization methodology to identify and then minimize linear, nonlinear, and interactive effects that induce covariate imbalance between groups. Investigators should consider using the CTA approach as a real-time complement to randomization for any clinical trial to safeguard the treatment allocation process against bias., (© 2017 John Wiley & Sons, Ltd.)

The exponential growth of digital text documents presents a significant challenge for text classification algorithms, as the vast number of words in each document can hinder their efficiency. Feature selection (FS) is a crucial technique that aims to eliminate irrelevant features and enhance classification accuracy. In this study, we propose an improved version of the discrete laying chicken algorithm (IDLCA) that utilizes noun‐based filtering to reduce the number of features and improve text classification performance. Although LCA is a newly proposed algorithm, it has not been systematically applied to discrete problems before. Our enhanced version of LCA employs different operators to improve both exploration and exploitation of this algorithm to find better solutions in discrete mode. To evaluate the effectiveness of the proposed method, we compared it with some conventional nature‐inspired feature selection methods using various learning models such as decision trees (DT), K‐nearest neighbor (KNN), Naive Bayes (NB), and support vector machine (SVM) on five benchmark datasets with three different evaluation metrics. The experimental results demonstrate the effectiveness of the proposed algorithm in comparison to the existing one. The code is available at https://github.com/m0javad/Improved-Discrete-Laying-Chicken-Algorithm. [ABSTRACT FROM AUTHOR]

This paper reports on progress in the development of a teaching module on machine learning with decision trees for secondary‐school students, in which students use survey data about media use to predict who plays online games frequently. This context is familiar to students and provides a link between school and everyday experience. In this module, they use CODAP's "Arbor" plug‐in to manually build decision trees and understand how to systematically build trees based on data. Further on, the students use a menu‐based environment in a Jupyter Notebook to apply an algorithm that automatically generates decision trees and to evaluate and optimize the performance of these. Students acquire technical and conceptual skills but also reflect on personal and social aspects of the uses of algorithms from machine learning. [ABSTRACT FROM AUTHOR]

Personalized medicine aims at identifying best treatments for a patient with given characteristics. It has been shown in the literature that these methods can lead to great improvements in medicine compared to traditional methods prescribing the same treatment to all patients. Subgroup identification is a branch of personalized medicine, which aims at finding subgroups of the patients with similar characteristics for which some of the investigated treatments have a better effect than the other treatments. A number of approaches based on decision trees have been proposed to identify such subgroups, but most of them focus on two-arm trials (control/treatment) while a few methods consider quantitative treatments (defined by the dose). However, no subgroup identification method exists that can predict the best treatments in a scenario with a categorical set of treatments. We propose a novel method for subgroup identification in categorical treatment scenarios. This method outputs a decision tree showing the probabilities of a given treatment being the best for a given group of patients as well as labels showing the possible best treatments. The method is implemented in an R package psica available on CRAN. In addition to a simulation study, we present an analysis of a community-based nutrition intervention trial that justifies the validity of our method. [ABSTRACT FROM AUTHOR]

Even though it is well known that for most relevant computational problems, different algorithms may perform better on different classes of problem instances, most researchers still focus on determining a single best algorithmic configuration based on aggregate results such as the average. In this paper, we propose integer programming‐based approaches to build decision trees for the algorithm selection problem. These techniques allow the automation of three crucial decisions: (a) discerning the most important problem features to determine problem classes, (b) grouping the problems into classes, and (c) selecting the best algorithm configuration for each class. To evaluate this new approach, extensive computational experiments were executed using the linear programming algorithms implemented in the COIN‐OR branch‐and‐cut solver across a comprehensive set of instances, including all MIPLIB benchmark instances. The results exceeded our expectations. While selecting the single best parameter setting across all instances decreased the total running time by 22%, our approach decreased the total running time by 40% on average across 10‐fold cross‐validation experiments. These results indicate that our method generalizes quite well and does not overfit. [ABSTRACT FROM AUTHOR]

Rationale, Aims and Objectives: The development of clinical practice guidelines (CPG) suffers from the lack of an explicit and transparent framework for synthesising the key elements necessary to formulate practice recommendations. We matched deliberations of the American Society of Haematology (ASH) CPG panel for the management of pulmonary embolism (PE) with the corresponding decision‐theoretical constructs to assess agreement of the panel recommendations with explicit decision modelling. Methods: Five constructs were identified of which three were used to reformulate the panel's recommendations: (1) standard, expected utility threshold (EUT) decision model; (2) acceptable regret threshold model (ARg) to determine the frequency of tolerable false negative (FN) or false positive (FP) recommendations, and (3) fast‐and‐frugal tree (FFT) decision trees to formulate the entire strategy for management of PE. We compared four management strategies: withhold testing versus d‐dimer → computerized pulmonary angiography (CTPA) ('ASH‐Low') versus CTPA→ d‐dimer ('ASH‐High') versus treat without testing. Results: Different models generated different recommendations. For example, according to EUT, testing should be withheld for prior probability PE < 0.13%, a clinically untenable threshold which is up to 15 times (2/0.13) below the ASH guidelines threshold of ruling out PE (at post probability of PE ≤ 2%). Three models only agreed that the 'ASH low' strategy should be used for the range of pretest probabilities of PE between 0.13% and 13.27% and that the 'ASH high' management should be employed in a narrow range of the prior PE probabilities between 90.85% and 93.07%. For all other prior probabilities of PE, choosing one model did not ensure coherence with other models. Conclusions: CPG panels rely on various decision‐theoretical strategies to develop its recommendations. Decomposing CPG panels' deliberation can provide insights if the panels' deliberation retains a necessary coherence in developing guidelines. CPG recommendations often do not agree with the EUT decision analysis, widely used in medical decision‐making modelling. [ABSTRACT FROM AUTHOR]

Abstract: Background: Contemporary delivery of health care is inappropriate in many ways, largely due to suboptimal Q5 decision‐making. A typical approach to improve practitioners' decision‐making is to develop evidence‐based clinical practice guidelines (CPG) by guidelines panels, who are instructed to use their judgments to derive practice recommendations. However, mechanisms for the formulation of guideline judgments remains a “black‐box” operation—a process with defined inputs and outputs but without sufficient knowledge of its internal workings. Methods: Increased explicitness and transparency in the process can be achieved by implementing CPG as clinical pathways (CPs) (also known as clinical algorithms or flow‐charts). However, clinical recommendations thus derived are typically ad hoc and developed by experts in a theory‐free environment. As any recommendation can be right (true positive or negative), or wrong (false positive or negative), the lack of theoretical structure precludes the quantitative assessment of the management strategies recommended by CPGs/CPs. Results: To realize the full potential of CPGs/CPs, they need to be placed on more solid theoretical grounds. We believe this potential can be best realized by converting CPGs/CPs within the heuristic theory of decision‐making, often implemented as fast‐and‐frugal (FFT) decision trees. This is possible because FFT heuristic strategy of decision‐making can be linked to signal detection theory, evidence accumulation theory, and a threshold model of decision‐making, which, in turn, allows quantitative analysis of the accuracy of clinical management strategies. Conclusions: Fast‐and‐frugal provides a simple and transparent, yet solid and robust, methodological framework connecting decision science to clinical care, a sorely needed missing link between CPGs/CPs and patient outcomes. We therefore advocate that all guidelines panels express their recommendations as CPs, which in turn should be converted into FFTs to guide clinical care. [ABSTRACT FROM AUTHOR]

The presented study applies the methods of data mining and prediction models to the subject of memory functioning in developmental dyslexia. This article sets forth the results of an analysis of the decision tree algorithm for the classification of dyslexia/non-dyslexia, based on frequency data from the modified simplified conjoint recognition experiment-a paradigm based on the fuzzy-trace theory used to investigate verbatim and gist memory. This decision tree model was created with the use of the C&RT algorithm, which makes a prediction of the classification with the use of four predictors: the numbers of different types of answers depending on the specific stimuli presented. Seventy-one high school students, 33 with developmental dyslexia, took part in a memory experiment. The model created using the decision tree algorithm has a very good overall validity. Excellent developmental dyslexia classification was accompanied by satisfactory non-dyslexia classification. The decision tree proposed predictors that are supported both theoretically and empirically. The results obtained show an important role of verbatim and gist memory functioning in developmental dyslexia and suggest that the pattern of performance observed in the memory tests can be used as a predictor of the developmental dyslexia disorder. Results encourage further usage of decision trees. [ABSTRACT FROM AUTHOR]

The family of decision tree learning algorithms is among the most widespread and studied. Motivated by the desire to develop learning algorithms that can generalize when learning highly varying functions such as those presumably needed to achieve artificial intelligence, we study some theoretical limitations of decision trees. We demonstrate formally that they can be seriously hurt by the curse of dimensionality in a sense that is a bit different from other nonparametric statistical methods, but most importantly, that they cannot generalize to variations not seen in the training set. This is because a decision tree creates a partition of the input space and needs at least one example in each of the regions associated with a leaf to make a sensible prediction in that region. A better understanding of the fundamental reasons for this limitation suggests that one should use forests or even deeper architectures instead of trees, which provide a form of distributed representation and can generalize to variations not encountered in the training data. [ABSTRACT FROM AUTHOR]

Objective: Body mass index (BMI) is the primary criterion differentiating anorexia nervosa (AN) and atypical anorexia nervosa despite prior literature indicating few differences between disorders. Machine learning (ML) classification provides us an efficient means of accurately distinguishing between two meaningful classes given any number of features. The aim of the present study was to determine if ML algorithms can accurately distinguish AN and atypical AN given an ensemble of features excluding BMI, and if not, if the inclusion of BMI enables ML to accurately classify between the two. Methods: Using an aggregate sample from seven studies consisting of individuals with AN and atypical AN who completed baseline questionnaires (N = 448), we used logistic regression, decision tree, and random forest ML classification models each trained on two datasets, one containing demographic, eating disorder, and comorbid features without BMI, and one retaining all features and BMI. Results: Model performance for all algorithms trained with BMI as a feature was deemed acceptable (mean accuracy = 74.98%, mean area under the receiving operating characteristics curve [AUC] = 74.75%), whereas model performance diminished without BMI (mean accuracy = 59.37%, mean AUC = 59.98%). Discussion: Model performance was acceptable, but not strong, if BMI was included as a feature; no other features meaningfully improved classification. When BMI was excluded, ML algorithms performed poorly at classifying cases of AN and atypical AN when considering other demographic and clinical characteristics. Results suggest a reconceptualization of atypical AN should be considered. Public Significance: There is a growing debate about the differences between anorexia nervosa and atypical anorexia nervosa as their diagnostic differentiation relies on BMI despite being similar otherwise. We aimed to see if machine learning could distinguish between the two disorders and found accurate classification only if BMI was used as a feature. This finding calls into question the need to differentiate between the two disorders. [ABSTRACT FROM AUTHOR]

Artificial intelligence has been showing accelerated growth due to its use in solving problems in several application domains. This success results from the convergence of large amounts of data, high‐performance computing, and precision of machine learning (ML) algorithms. Even with the relevance of ML algorithms, little is known about their computational requirements and power consumption, which has become an important task to achieve greener computing. This work aims to evaluate the energy efficiency of the ML algorithms to identify their energy hotspots. Also, to investigate which influences the energy consumption (EC) of these algorithms and how the parameters design could affect it. We conducted a series of experiments using 27 different datasets, 2 decision trees algorithms, and 2 ensembles for classification and regression tasks to answer these questions. Our results show interesting findings, such as, like some simple parameters choice can have a high impact on EC, as a consequence, finding greener strategies for AI. [ABSTRACT FROM AUTHOR]

Aim: To develop and validate models based on logistic regression and artificial intelligence for prognostic prediction of molar survival in periodontally affected patients. Materials and Methods: Clinical and radiographic data from four different centres across four continents (two in Europe, one in the United States, and one in China) including 515 patients and 3157 molars were collected and used to train and test different types of machine‐learning algorithms for their prognostic ability of molar loss over 10 years. The following models were trained: logistic regression, support vector machine, K‐nearest neighbours, decision tree, random forest, artificial neural network, gradient boosting, and naive Bayes. In addition, different models were aggregated by means of the ensembled stacking method. The primary outcome of the study was related to the prediction of overall molar loss (MLO) in patients after active periodontal treatment. Results: The general performance in the external validation settings (aggregating three cohorts) revealed that the ensembled model, which combined neural network and logistic regression, showed the best performance among the different models for the prediction of MLO with an area under the curve (AUC) = 0.726. The neural network model showed the best AUC of 0.724 for the prediction of periodontitis‐related molar loss. In addition, the ensembled model showed the best calibration performance. Conclusions: Through a multi‐centre collaboration, both prognostic models for the prediction of molar loss were developed and externally validated. The ensembled model showed the best performance in terms of both discrimination and validation, and it is made freely available to clinicians for widespread use in clinical practice. [ABSTRACT FROM AUTHOR]

Rationale, Aims and Objectives: Critics have charged that evidence‐based medicine (EBM) overemphasises algorithmic rules over unstructured clinical experience and intuition, but the role of structured decision support systems in improving health outcomes remains uncertain. We aim to assess if delivery of anticoagulant prophylaxis in hospitalised patients with COVID‐19 according to an algorithm based on evidence‐based clinical practice guideline (CPG) improved clinical outcomes compared with administration of anticoagulant treatment given at individual practitioners' discretion. Methods: An observational design consisting of the analysis of all acutely ill, consecutive patients (n = 1783) with confirmed COVID‐19 diagnosis admitted between 10 March 2020 to 11 January 2022 to an US academic center. American Society of Haematology CPG for anticoagulant prophylaxis in hospitalised patients with COVID‐19 was converted into a clinical pathway and translated into fast‐and‐frugal decision (FFT) tree ('algorithm'). We compared delivery of anticoagulant prophylaxis in hospitalised patients with COVID‐19 according to the FFT algorithm with administration of anticoagulant treatment given at individual practitioners' discretion. Results: In an adjusted analysis, using combination of Lasso (least absolute shrinkage and selection operator) and propensity score based weighting [augmented inverse‐probability weighting] statistical techniques controlling for cluster data, the algorithm did not reduce death, venous thromboembolism, or major bleeding, but helped avoid longer hospital stay [number of patients needed to be treated (NNT) = 40 (95% CI: 23–143), indicating that for every 40 patients (23–143) managed on FFT algorithm, one avoided staying in hospital longer than 10 days] and averted admission to intensive‐care unit (ICU) [NNT = 19 (95% CI: 13–40)]. All model's selected covariates were well balanced. The results remained robust to sensitivity analyses used to test the stability of the findings. Conclusions: When delivered using a structured FFT algorithm, CPG shortened the hospital stay and help avoided admission to ICU, but it did not affect other relevant outcomes. [ABSTRACT FROM AUTHOR]

Background: This study aimed to investigate the value of nonenhanced computed tomography (CT)‐based radiomics in determining disease progression in breast cancer patients with bone marrow metastases and to develop a model for assessing treatment efficacy. Methods: A total of 134 breast cancer patients with bone metastases were enrolled from three hospitals. Nonenhanced CT was performed after two cycles of drug treatment. The images were categorized into an invalid and a valid group according to disease progression status. The largest osteolytic lesions' maximum cross‐sections in the CT images were selected as regions of interest (ROIs) for feature extraction. Variance threshold, SelectKBest, and least absolute shrinkage and selection operator (LASSO) were used to reduce feature dimensionality. K‐nearest neighbor algorithm (KNN), support vector machine (SVM), extreme gradient boosting (XGBoost), random forest (RF), logistic regression (LR), and decision tree (DT) algorithms were trained to establish radiomics models. Receiver operating characteristic (ROC) curves were generated to evaluate the diagnostic performance of the models. Results: The KNN classifier demonstrated the best performance compared to the random grouping method. In the validation group, the area under the ROC curve (AUC) was 0.810. In the cross‐validation method, the RF classifier showed the best performance with an AUC of 0.84. Conclusion: Nonenhanced CT‐based radiomics provides a promising method for evaluating the efficacy of systemic drug therapy in breast cancer patients with osteolytic bone metastases. [ABSTRACT FROM AUTHOR]

There is vigorous debate as to whether influential social media platforms like Twitter and Facebook should censor objectionable posts by prominent individuals in the United States and elsewhere. A tentative middle ground is employing content moderation to signal to social media audiences that certain posts may contain objectionable information through the mechanism of flagging. Existing studies have mainly examined the effect of flagging regular users' content. To add to this emerging literature stream, we examine the effect of flagging when the underlying content is produced by a prominent individual. Leveraging Twitter's moderation activities on former U.S. President Donald Trump's tweets as our empirical setting, we employ three machine learning algorithms to estimate the effect of flagging Trump's tweets. We explore preliminary evidence as to whether these posts were retweeted less or more than expected. Our results indicate that the flagged tweets were retweeted at higher rates than expected. Our findings suggest that flagging content of prominent individuals on social media might be ineffective or even counterproductive in curbing the spread of content deemed objectionable by social media companies. [ABSTRACT FROM AUTHOR]

Background: Machine learning investigates how computers can automatically learn. The present study aimed to predict dietary patterns and compare algorithm performance in making predictions of dietary patterns. Methods: We analysed the data of public employees (n = 12,667) participating in the Brazilian Longitudinal Study of Adult Health (ELSA‐Brasil). The K‐means clustering algorithm and six other classifiers (support vector machines, naïve Bayes, K‐nearest neighbours, decision tree, random forest and xgboost) were used to predict the dietary patterns. Results: K‐means clustering identified two dietary patterns. Cluster 1, labelled the Western pattern, was characterised by a higher energy intake and consumption of refined cereals, beans and other legumes, tubers, pasta, processed and red meats, high‐fat milk and dairy products, and sugary beverages; Cluster 2, labelled the Prudent pattern, was characterised by higher intakes of fruit, vegetables, whole cereals, white meats, and milk and reduced‐fat milk derivatives. The most important predictors were age, sex, per capita income, education level and physical activity. The accuracy of the models varied from moderate to good (69%–72%). Conclusions: The performance of the algorithms in dietary pattern prediction was similar, and the models presented may provide support in screener tasks and guide health professionals in the analysis of dietary data. Key points: Machine learning (ML) investigates how computers can automatically learn. The present study aimed to predict dietary patterns and to compare the performance of various ML algorithms for making the predictions of dietary patterns.K‐means clustering identified two major dietary patterns. The models presented may provide support in screener tasks. [ABSTRACT FROM AUTHOR]

LEarning TO Rank (LETOR) algorithms are usually trained on annotated corpora where a single relevance label is assigned to each available document‐topic pair. Within the Cranfield framework, relevance labels result from merging either multiple expertly curated or crowdsourced human assessments. In this paper, we explore how to train LETOR models with relevance judgments distributions (either real or synthetically generated) assigned to document‐topic pairs instead of single‐valued relevance labels. We propose five new probabilistic loss functions to deal with the higher expressive power provided by relevance judgments distributions and show how they can be applied both to neural and gradient boosting machine (GBM) architectures. Moreover, we show how training a LETOR model on a sampled version of the relevance judgments from certain probability distributions can improve its performance when relying either on traditional or probabilistic loss functions. Finally, we validate our hypothesis on real‐world crowdsourced relevance judgments distributions. Overall, we observe that relying on relevance judgments distributions to train different LETOR models can boost their performance and even outperform strong baselines such as LambdaMART on several test collections. [ABSTRACT FROM AUTHOR]

Background and aims: Phosphatidylethanol 16:0/18:1 (PEth) is a biomarker for alcohol intake. It has a half‐life of 7.9 days. Chronic alcohol consumption causes high PEth values. It can take weeks before PEth values fall below the decision limit for 'alcohol abstinence'. Our aim was to validate whether alcohol abstinence can be determined based on two consecutive PEth results above the decision limit. Design Observational study. Setting: Belgium, February 2019. The study was linked to a social initiative in Belgium, 'Tournée Minérale'. Participants: Adults (aged > 18 years, n = 796) with varying drinking habits who self‐reportedly refrained from alcohol consumption during the study. Measurements A validated liquid chromatography–tandem mass spectrometry method was used to quantify PEth in participants' dried blood samples, collected at three time‐points via remote fingerprick‐based self‐sampling. Findings A population‐based algorithm to evaluate abstinence based on 95% prediction limits was developed by fitting a linear mixed‐effect model to discern patterns in PEth elimination over time. It took intra‐ and inter‐individual variability into consideration. The algorithm was included in a two‐step decision tree, assessing whether (i) PEth values fell within the prediction interval and (ii) the slope between two PEth values was consistent with no alcohol consumption. Data for 74 participants reporting no alcohol intake during the study were used for validation. With a detection limit of 'four units spread over 14 days', the sensitivity and specificity of the decision tree was 89%. Conclusions: Claims of alcohol abstinence can be verified using a two‐step decision tree for phosphatidylethanol 16:0/18:1 values, even when those values are above the limit for 'alcohol abstinence'. [ABSTRACT FROM AUTHOR]

Aims: Automated checks for medication-related problems have become a cornerstone of medication safety. In many clinical settings medication checks remain confined to drug-drug interactions because only medication data are available in an adequately coded form, leaving possible contraindicated drug-disease combinations unaccounted for. Therefore, we devised algorithms that identify frequently contraindicated diagnoses based on medication patterns related to these diagnoses., Methods: We identified drugs that are associated with diseases constituting common contraindications based on their exclusive use for these conditions (such as allopurinol for gout or salbutamol for bronchial obstruction). Expert-based and machine learning algorithms were developed to identify diagnoses based on highly specific medication patterns. The applicability, sensitivity and specificity of the approach were assessed by using an anonymized real-life sample of medication and diagnosis data excerpts from 3506 discharge records of geriatric patients., Results: Depending on the algorithm, the desired focus (i.e., sensitivity vs. specificity) and the disease, we were able to identify the diagnoses gout, epilepsy, coronary artery disease, congestive heart failure and bronchial obstruction with a specificity of 44.0-99.8% (95% CI 41.7-100.0%) and a sensitivity of 3.8-83.1% (95% CI 1.0-86.1%). Using only medication data, we were able to identify 123 (51.3%) of 240 contraindications identified by experts with access to medication data and diagnoses., Conclusion: This study provides a proof of principle that some key diagnosis-related contraindications can be identified based on a patient's medication data alone, while others cannot be identified. This approach offers new opportunities to analyse drug-disease contraindications in community pharmacy or clinical routine data., (© 2022 The Authors. British Journal of Clinical Pharmacology published by John Wiley & Sons Ltd on behalf of British Pharmacological Society.)

Problem: Autism spectrum disorders (ASD) are pervasive neurodevelopmental disorders and generally accompanied by social disorders, verbal or nonverbal communication defects, inability to concentrate and other negative symptoms that affect the autistic person's normal life. However, traditional screening methods are time‐consuming and public health resources are limited. Methods: This study proposed a novel technique that combined eye‐movement data and machine learning algorithms for predicting autistic traits. We converted raw eye movement data into features, trained and tested a model for early screening. Findings: In the preliminary experiment, 107 participants (average age = 24.84 ± 5.24 years) wore HTC Vive to watch a VR scene for 15–20 s. We explored eight classification models, among which the ensemble model performed best, with 0.73 accuracy, 0.68 precision, 0.81 recall, 0.74 F1‐score, and an area under the curve of 0.90. And in the test experiment, 22 participants (average age = 12.68 ± 7.61 years) diagnosed as ASD took the experiment and the ensemble model showed a recall of 0.77. Eye movement data is an effectively distinguishable tool and we find that the proportion of time to observe figure and animal region continuously can distinguish participants with obvious and unobvious autistic traits effectively in the model. Conclusion: This study focuses on the detection of autistic traits, and proposes a more objective and faster method for undertaking early screening, which provides possibilities to save precious time to intervene and alleviate its symptoms before making a definite diagnosis. [ABSTRACT FROM AUTHOR]

Premise: Advancements in machine learning and the rise of accessible "big data" provide an important opportunity to improve trait‐based plant identification. Here, we applied decision‐tree induction to a subset of data from the TRY plant trait database to (1) assess the potential of decision trees for plant identification and (2) determine informative traits for distinguishing taxa. Methods: Decision trees were induced using 16 vegetative and floral traits (689 species, 20 genera). We assessed how well the algorithm classified species from test data and pinpointed those traits that were important for identification across diverse taxa. Results: The unpruned tree correctly placed 98% of the species in our data set into genera, indicating its promise for distinguishing among the species used to construct them. Furthermore, in the pruned tree, an average of 89% of the species from the test data sets were properly classified into their genera, demonstrating the flexibility of decision trees to also classify new species into genera within the tree. Closer inspection revealed that seven of the 16 traits were sufficient for the classification, and these traits yielded approximately two times more initial information gain than those not included. Discussion: Our findings demonstrate the potential for tree‐based machine learning and big data in distinguishing among taxa and determining which traits are important for plant identification. [ABSTRACT FROM AUTHOR]

Introduction: Diagnosis of outflow tract ventricular arrhythmia (OTVA) localization by an electrocardiographic complex is key to successful catheter ablation for OTVA. However, diagnosing the origin of OTVA with a precordial transition in lead V3 (V3TZ) is challenging. This study aimed to create the best practical electrocardiogram algorithm to differentiate the left ventricular outflow tract (LVOT) from the right ventricular outflow tract (RVOT) of OTVA origin with V3TZ using machine learning. Methods: Of 498 consecutive patients undergoing catheter ablation for OTVA, we included 104 patients who underwent ablation for OTVA with V3TZ and identified the origin of LVOT (n = 62) and RVOT (n = 42) from the results. We analyzed the standard 12‐lead electrocardiogram preoperatively and measured 128 elements in each case. The study population was randomly divided into training group (70%) and testing group (30%), and decision tree analysis was performed using the measured elements as features. The performance of the algorithm created in the training group was verified in the testing group. Results: Four measurements were identified as important features: the aVF/II R‐wave ratio, the V2S/V3R index, the QRS amplitude in lead V3, and the R‐wave deflection slope in lead V3. Among them, the aVF/II R‐wave ratio and the V2S/V3R index had a particularly strong influence on the algorithm. The performance of this algorithm was extremely high, with an accuracy of 94.4%, precision of 91.5%, recall of 100%, and an F1‐score of 0.96. Conclusions: The novel algorithm created using machine learning is useful in diagnosing the origin of OTVA with V3TZ. [ABSTRACT FROM AUTHOR]

Background: Considering the high morbidity and mortality of lung cancer and the high incidence of pulmonary nodules, clearly distinguishing benign from malignant lung nodules at an early stage is of great significance. However, determining the kind of lung nodule which is more prone to lung cancer remains a problem worldwide. Methods: A total of 480 patients with pulmonary nodule data were collected from Shandong, China. We assessed the clinical characteristics and computed tomography (CT) imaging features among pulmonary nodules in patients who had undergone video‐assisted thoracoscopic surgery (VATS) lobectomy from 2013 to 2018. Preliminary selection of features was based on a statistical analysis using SPSS. We used WEKA to assess the machine learning models using its multiple algorithms and selected the best decision tree model using its optimization algorithm. Results: The combination of decision tree and logistics regression optimized the decision tree without affecting its AUC. The decision tree structure showed that lobulation was the most important feature, followed by spiculation, vessel convergence sign, nodule type, satellite nodule, nodule size and age of patient. Conclusions: Our study shows that decision tree analyses can be applied to screen individuals for early lung cancer with CT. Our decision tree provides a new way to help clinicians establish a logical diagnosis by a stepwise progression method, but still needs to be validated for prospective trials in a larger patient population. [ABSTRACT FROM AUTHOR]

Liver cancer has become the third cause that leads to the cancer death. For hepatocellular carcinoma (HCC), as the highly malignant type of liver cancer, its recurrence rate after operation is still very high because there is no reliable clinical data to provide better advice for patients after operation. To solve the challenging issue, in this work, we design a novel prediction model for recurrence of HCC using neighbor2vec based algorithm. It consists of three stages: (a) In the preparation stage, the Pearson correlation coefficient was used to explore the independent predictors of HCC recurrence, (b) due to the low correlation between individual dimension and prediction target, K‐nearest neighbors (KNN) were found as a K‐vectors list for each patient (neighbor2vec), (c) all vectors lists were applied as the input of machine learning methods such as logistic regression, KNN, decision tree, naive Bayes (NB), and deep neural network to establish the neighbor2vec based prediction model. From the experimental results on the real data from Shandong Provincial Hospital in China, the proposed neighbor2vec based prediction model outperforms all the other models. Especially, the NB model with neighbor2vec achieves up to 83.02, 82.86, 77.6%, in terms of accuracy, recall rates, and precision. This article is categorized under:Technologies > Data PreprocessingTechnologies > ClassificationTechnologies > Machine Learning [ABSTRACT FROM AUTHOR]

This paper proposes the model and design of optimal utilization of grid connected Micro Grid (MG) system with Enhanced Fruit fly Optimization with Gradient Boosting Decision Tree (EFO‐GBDT) based controller technique. The integration (modeling and design) of high performance converter topology is formulated with the consideration of enhanced converter efficiency, effective implementation, source of renewable energy and reduced swapping loss. In the proposed technique the control scheme is used to consolidate the performance of both an EFO algorithm and also GBDT. In the proposed technique, the EFOA plays out an estimate method to initiate an exact control signal systems and accumulates the control signal database, an offline way of light in power variety between the source side (source of renewable energy and battery) and load side. The performance of dataset used to work as the GBDT mode for an online way and it leads to the control procedure in less performance time. In this proposed technique, an objective function is determined by the system, data subject to be an equality and inequality restrictions. The restrictions are the sources of renewable energy, power require the storage of an charged elements. A battery is implemented as source energy and it is used to permit and balance the renewable power system units and to keep running at a steady and balance output power. At that point the proposed model is performed based on MATLAB/Simulink, its working platform and the performance is assessed based on the existing techniques. [ABSTRACT FROM AUTHOR]

Features selection is the process of choosing the relevant subset of features from the high‐dimensional dataset to enhance the performance of the classifier. Much research has been carried out in the present world for the process of feature selection. Algorithms such as Naïve Bayes (NB), decision tree, and genetic algorithm are applied to the high‐dimensional dataset to select the relevant features and also to increase the computational speed. The proposed model presents a solution for selection of features using ensemble classifier algorithms. The proposed algorithm is the combination of minimum redundancy and maximum relevance (mRMR) and forest optimization algorithm (FOA). Ensemble‐based algorithms such as support vector machine (SVM), K‐nearest neighbor (KNN), and NB is further used to enhance the performance of the classifier algorithm. The mRMR‐FOA is used to select the relevant features from the various datasets and 21% to 24% improvement is recorded in the feature selection. The ensemble classifier algorithms further improves the performance of the algorithm and provides accuracy of 96%. [ABSTRACT FROM AUTHOR]

An object-based evaluation method using a pattern recognition algorithm (i.e., classification trees) is applied to the simulated orographic precipitation for idealized experimental setups using the National Center of Atmospheric Research (NCAR) Community Atmosphere Model (CAM) with the finite volume (FV) and the Eulerian spectral transform dynamical cores with varying resolutions. Daily simulations were analyzed and three different types of precipitation features were identified by the classification tree algorithm. The statistical characteristics of these features (i.e., maximum value, mean value, and variance) were calculated to quantify the difference between the dynamical cores and changing resolutions. Even with the simple and smooth topography in the idealized setups, complexity in the precipitation fields simulated by the models develops quickly. The classification tree algorithm using objective thresholding successfully detected different types of precipitation features even as the complexity of the precipitation field increased. The results show that the complexity and the bias introduced in small-scale phenomena due to the spectral transform method of CAM Eulerian spectral dynamical core is prominent, and is an important reason for its dissimilarity from the FV dynamical core. The resolvable scales, both in horizontal and vertical dimensions, have significant effect on the simulation of precipitation. The results of this study also suggest that an efficient and informative study about the biases produced by GCMs should involve daily (or even hourly) output (rather than monthly mean) analysis over local scales. [ABSTRACT FROM AUTHOR]

Objectives: Health services have an imperative to reduce prolonged patient length of stay (LOS) in ED. Our objective is to develop and validate an accurate prediction model for patient LOS in ED greater than 4 hours using a data mining technique. Methods: Data were collected from a regional Australian public hospital for all ED presentations between 1 January 2016 and 31 December 2017. A decision tree algorithm was built to predict patients with an ED LOS >4 hours. A total of 33 attributes were analysed. The performance of the final model was internally validated. Clinically relevant patterns from the model were analysed. Results: The accuracy of the model was 85%. We identified that patients at our site who were at high risk of ED LOS >4 hours were those who were waiting in ED for a medical consultation, or those who were waiting for a urology, surgical, orthopaedic or paediatric consultation if the request for consultation occurred more than 2 hours after the patient was first seen by an ED doctor. Conclusion: This model performed very well in predicting ED LOS >4 hours for each individual patient and demonstrated a number of clinically relevant patterns. Identifying patterns that influence ED LOS is important for health managers in order to develop and implement interventions targeted at those clinical scenarios. Future work should look at the utility of displaying individual patient risk of ED LOS >4 hours using this model in real‐time at the point‐of‐care. [ABSTRACT FROM AUTHOR]

This article addresses the problem of detecting crisis‐related messages on social media, in order to improve the situational awareness of emergency services. Previous work focused on developing machine‐learning classifiers restricted to specific disasters, such as storms or wildfires. We investigate for the first time methods to detect such messages where the type of the crisis is not known in advance, that is, the data are highly heterogeneous. Data heterogeneity causes significant difficulties for learning algorithms to generalize and accurately label incoming data. Our main contributions are as follows. First, we evaluate the extent of this problem in the context of disaster management, finding that the performance of traditional learners drops by up to 40% when trained and tested on heterogeneous data vis‐á‐vis homogeneous data. Then, in order to overcome data heterogeneity, we propose a new ensemble learning method, and found this to perform on a par with the Gradient Boosting and AdaBoost ensemble learners. The methods are studied on a benchmark data set comprising 26 disaster events and four classification problems: detection of relevant messages, informative messages, eyewitness reports, and topical classification of messages. Finally, in a case study, we evaluate the proposed methods on a real‐world data set to assess its practical value. [ABSTRACT FROM AUTHOR]

The quality of a cluster analysis of unlabeled units depends on the quality of the between units dissimilarity measures. Data‐dependent dissimilarity is more objective than data independent geometric measures such as Euclidean distance. As suggested by Breiman, many data driven approaches are based on decision tree ensembles, such as a random forest (RF), that produce a proximity matrix that can easily be transformed into a dissimilarity matrix. An RF can be obtained using labels that distinguish units with real data from units with synthetic data. The resulting dissimilarity matrix is input to a clustering program and units are assigned labels corresponding to cluster membership. We introduce a general iterative cluster (GIC) algorithm that improves the proximity matrix and clusters of the base RF. The cluster labels are used to grow a new RF yielding an updated proximity matrix, which is entered into the clustering program. The process is repeated until convergence. The same procedure can be used with many base procedures such as the extremely randomized tree ensemble. We evaluate the performance of the GIC algorithm using benchmark and simulated data sets. The properties measured by the Silhouette score are substantially superior to the base clustering algorithm. The GIC package has been released in R: https://cran.r‐project.org/web/packages/GIC/index.html. [ABSTRACT FROM AUTHOR]

Background: To date, incident and recurrent MI remains a major health issue worldwide, and efforts to improve risk prediction in population health studies are needed. This may help the scalability of prevention strategies and management in terms of healthcare cost savings and improved quality of care. Methods: We studied a large‐scale population of 4.3 million US patients from different socio‐economic and geographical areas from three health plans (Commercial, Medicare, Medicaid). Individuals had medical/pharmacy benefits for at least 30 months (2 years for comorbid history and followed up for 6 months or more for clinical outcomes). Machine‐learning (ML) algorithms included supervised (logistic regression, neural network) and unsupervised (decision tree, gradient boosting) methodologies. Model discriminant validity, calibration and clinical utility were performed separately on allocated test sample (1/3 of original data). Results: In the absence of MI in comorbid history, the overall incidence rates were 0.442 cases/100 person‐years and in the presence of MI history, 0.652. ML algorithms showed that supervised formulations had incrementally higher discriminant validity than unsupervised techniques (e.g., for incident MI outcome in the absence of MI in comorbid history: logistic regression "LR" – c index 0.921, 95%CI 0.920–0.922; neural network "NN" – c index 0.914, 95%CI 0.913–0.915; gradient boosting "GB" – c index 0.902, 95%CI 0.900–0.904; decision tree "DT" – c index 0.500, 95%CI 0.495–0.505). Calibration and clinical utility showed good to excellent results. Conclusion: ML algorithms can substantially improve the prediction of incident and recurrent MI particularly in terms of the non‐linear formulation. This approach may help with improved risk prediction, allowing implementation of cardiovascular prevention strategies across diversified sub‐populations with different clusters of complexity. [ABSTRACT FROM AUTHOR]

SMS, one of the most popular and fast‐growing GSM value‐added services worldwide, has attracted unwanted SMS, also known as SMS spam. The effects of SMS spam are significant as it affects both the users and the service providers, causing a massive gap in trust among both parties. This article presents a deep learning model based on BiLSTM. Further, it compares our results with some of the states of the art machine learning (ML) algorithm on two datasets: our newly collected dataset and the popular UCI SMS dataset. This study aims to evaluate the performance of diverse learning models and compare the result of the new dataset expanded (ExAIS_SMS) using the following metrics the true positive (TP), false positive (FP), F‐measure, recall, precision, and overall accuracy. The average accuracy for the BiLSTSM model achieved moderately improved results compared to some of the ML classifiers. The experimental results achieved significant improvement from the ground truth results after effective fine‐tuning of some of the parameters. The BiLSTM model using the ExAIS_SMS dataset attained an accuracy of 93.4% and 98.6% for UCI datasets. Further comparison of the two datasets on the state‐of‐the‐art ML classifiers gave an accuracy of Naive Bayes, BayesNet, SOM, decision tree, C4.5, J48 is 89.64%, 91.11%, 88.24%, 75.76%, 80.24%, and 79.2% respectively for ExAIS_SMS datasets. In conclusion, our proposed BiLSTM model showed significant improvement over traditional ML classifiers. To further validate the robustness of our model, we applied the UCI datasets, and our results showed optimal performance while classifying SMS spam messages based on some metrics: accuracy, precision, recall, and F‐measure. [ABSTRACT FROM AUTHOR]

Background: Merkel cell carcinoma (MCC) is a rare neuroendocrine skin cancer with a high mortality rate. MCC staging is currently based on tumour primary size, clinical detectability of lymph node metastases, performance of a lymph node biopsy, and presence of distant metastases., Objective: We aimed to use a modified classification and regression tree (CART) algorithm using available data points in the National Cancer Database (NCDB) to elucidate novel prognostic factors for MCC., Methods: Retrospective cohort study of the NCDB and Surveillance, Epidemiology, and End Results (SEER) registries. Cases from the NCDB were randomly assigned to either the training or validation cohorts. A modified CART algorithm was created with data from the training cohort and used to identify prognostic groups that were validated in the NCDB validation and SEER cohorts., Results: A modified CART algorithm using tumour variables available in the NCDB identified prognostic strata as follows: I: local disease, II: ≤3 positive nodes, III: ≥4 positive nodes, and IV: presence of distant metastases. Three-year survival for these groups in the NCDB validation cohort were 81.2% (SE: 1.7), 59.6% (SE: 3.0), 38.0% (SE: 6.0), and 20.2% (SE: 7.0), respectively. These strata were exhibited greater within-group homogeneity than AJCC groups and were more predictive of survival., Conclusions: Risk-stratified grouping of MCC patients incorporating positive lymph node count were strongly predictive of survival and demonstrated a high degree of within-group homogeneity and survival prediction. Incorporation of positive lymph node count within overall staging or sub-staging may help to improve future MCC staging criteria., (© 2021 The Australasian College of Dermatologists.)

Background: The present study aimed to evaluate the performance of several machine learning (ML) algorithms in predicting 1‐year afatinib continuation and 2‐year survival after afatinib initiation and to identify the differences in survival outcomes between ML‐classified strata. Methods: Data that were also used in the RESET study were retrospectively collected from 16 hospitals in South Korea. A stratified random sampling method was applied to split the data into training and test sets (70:30 split ratio). Clinical information, such as age, sex, tumor stage, smoking, performance status, metastasis, type of metastasis, dose adjustment, and pathologic information on EGFR mutations were inputted. Training was performed using eight ML algorithms: logistic regression, decision tree, deep neural network, random forest, support vector machine, boosting, bagging, and the naïve Bayes classifier. The model performance was assessed based on sensitivity, specificity, and accuracy. Area under the receiver operator characteristic curve (AUC) was calculated and compared between the ML models using DeLong's test. A Kaplan–Meier (KM) curve was used to visualize the identified strata obtained from the ML models. Results: No significant differences in the input variables were observed between the training and test datasets. The best‐performing models were support vector machine in predicting 1‐year afatinib continuation (AUC 0.626) and decision tree in 2‐year survival after afatinib start (AUC 0.644), although the performances of the ML models were comparable and did not display any predictive roles. KM analysis and log‐rank test revealed significant differences between the strata identified from the ML model (p < 0.001) in terms of both time‐on‐treatment (TOT) and overall survival (OS). Conclusion: The performances of ML models in our study found no discernible roles in predicting afatinib‐related outcomes, although the identified strata revealed different TOT and OS in the KM analysis. This implies the strength of ML in predicting the survival outcome, as well as the limitation of electronic medical record‐based variables in ML algorithms. Careful consideration of variable inclusion is likely to improve the general model performance. [ABSTRACT FROM AUTHOR]

Background and objective: Variations in the risk factors for sarcopenia can lead to differences in the likelihood of developing sarcopenia among older adults; however, few studies have explored the interactions among the risk factors. This study examined the interactions among risk factors and identified a discriminative pathway for groups at risk of sarcopenia in community‐dwelling older adults. Methods: A cross‐sectional study was conducted between July and August 2019 to recruit 200 older adults from an outpatient department of a hospital providing care for older people. Data on various risk factors, namely demographics (age, gender, education, comorbidities, and body mass index [BMI]), dietary habits (weekly consumption of milk, coffee, and meat), lifestyle behaviours (vitamin D supplementation, smoking, drinking, and physical activity), and depression symptoms were collected. Sarcopenia was defined according to the Asian Working Group for Sarcopenia criteria. A classification and regression tree (CART) model was used to examine interactions among these factors and identify groups at risk of sarcopenia. Findings: The prevalence of sarcopenia was 38.5%. The CART model identified two end groups at differential risks of sarcopenia, with a minimum of one and a maximum of three risk factors. In the first group, low BMI (<18.5 kg/m2) was a predominant risk factor for sarcopenia among older people. In the second group, older adults with a normal BMI, aged ≥68 years, and without a regular walking habit had a higher probability of developing sarcopenia than did their counterparts. Conclusions: The interactive effects among older age, BMI, and walking may cause different probabilities of developing sarcopenia in the older population. Implications for practice: Older adults with a low or normal BMI but without a regular walking habit could be a predominant risk group for sarcopenia. The appropriate maintenance of body weight and regular walking activity is suggested to prevent sarcopenia in community‐dwelling older adults. [ABSTRACT FROM AUTHOR]

Small‐pore zeolites possess pores that are constructed of eight tetrahedral atoms. ZSM‐43 is a small‐pore zeolite with two‐dimensional eight‐ring channels. The preparation of ZSM‐43 is influenced by the molar concentration of hydroxide ions, choline based organic structure‐directing agents (OSDA) and inorganic structure‐directing agents (ISDA) such as sodium, potassium and cesium. The synthetic conditions yield a range of products such as ZSM‐43, amorphous, UZM‐15, and other zeolites. There is a significant challenge in relating synthetic descriptors to their implications in zeolite phase formation. As a proof‐of‐concept study, we correlated the type of product formed with the gel molar composition in complex ZSM‐43 synthesis using machine learning algorithms. Seven different supervised machine learning algorithms have demonstrated an accuracy of >95 percent and an F1 score of >95 percent. The decision tree algorithm (DT) demonstrates the relationship between what type of phase is produced by what concentration of ISDA and hydroxide ions, as well as the effects of changing these parameters on phase transformation. DT provides structural and physicochemical insights into zeolite chemistry. From the experimental data of ZSM‐43, it is difficult to obtain detailed information regarding the chemistry of all phase formations. However, machine learning algorithms aid in recognizing hidden patterns in the data, facilitating deeper understanding of zeolite chemistry and phase transformations. [ABSTRACT FROM AUTHOR]

Research indicates that selecting a strategy to best exploit a new technology is a complex decision-making process. The task involves making a series of decisions with multiple alternatives, each to be evaluated by multiple criteria whose values have high levels of uncertainty. This paper presents a methodology for modelling a new technology decision using decision trees and an optimizing algorithm. A problem of a mining company considering the adoption of new technology is used to illustrate the decision-making task and modelling methodology. A numerical solution to the case demonstrates the potential of the optimizing technique in strategy selection. [ABSTRACT FROM AUTHOR]

The selection of the infiltration equation is important in various water research and management fields. Different infiltration equations were found to perform better than others in previous studies. Our objective was to use the Soil Water Infiltration Global (SWIG) database for (a) evaluating and comparing infiltration equations and (b) using the random forest algorithm to investigate how soil properties, land use, and infiltration measurement methods influence the infiltration equation selection. The performance of six equations (Horton, Mezencev, Collis‐George, Green–Ampt, Philip, and Swartzendruber) was characterized by the Akaike information criterion obtained after fitting them to 4,830 cumulative infiltration datasets from the SWIG database. Then, the random forest algorithm was applied to predict the best infiltration equation using soil texture class, organic matter content, bulk density, saturated hydraulic conductivity, land use, and infiltration measurement method as inputs. The Horton, Mezencev, and Collis‐George models were the best in 36, 24, and 12% of cases, respectively. Swartzendruber, Philip, and Green–Ampt were the best in 11, 10, and 7% cases, respectively. The different error of predicting the best infiltration equation was observed in different parts of input variable space. The infiltration method was by far the most important predictor for a model being the best across the whole database, followed by the organic C content and land use type. Organic C content and land use type were the most important predictors for the tension infiltrometer datasets. The SWIG database presents the opportunity to forecast which infiltration equation will work best in site‐specific conditions. Core Ideas: Selection of the infiltration equation is based on basic soil properties and land use.Data from the global SWIG database were fitted with six popular infiltration equations.The infiltration measurement method was the lead predictor for the equation selection.It was followed by the organic C content and the land use type.The error of the best infiltration equation predictions depends on soil properties and land use. [ABSTRACT FROM AUTHOR]

The problem of choosing strategies which minimize the cost of decision tree testing is considered in the case where the cost of each test depends on its outcome. The classical entropy bound on the cost is derived and we propose an algorithm for decision tree design based on an extension of Huffman's construction. By comparing the mean costs for a large sample of randomly chosen situations, we show that this algorithm performs well compared with a number of previously published algorithms. The algorithms are also compared with optimal performance for a range of simple cases. [ABSTRACT FROM AUTHOR]

Different nature‐inspired meta‐heuristic algorithms have been proposed to solve optimization problems. One of these algorithms is called social spider optimization (SSO) algorithm. Spiders' natural behaviors have inspired them to find the bait position by detecting vibrations in their web. Although the SSO algorithm has good accuracy in achieving optimal solutions, it suffers from a low convergence rate. In this paper, we attempted to improve SSO by changing its motion and mating parameters. To provide a practical example of using the new proposed algorithm, we based it on multi‐objective opposition‐based SSO, named MOPSSO. We used this algorithm in a feature selection process for analyzing text psychology, which is a multi‐objective problem. Textual psychology analysis is used in various fields, including collecting and analyzing people's views on various products, topics, social and political events. After selecting features, in order to classify the text, we used a new hybrid method that hybrids fuzzy C‐MEANS data clustering technique, a decision tree (DT), and Naïve Bayes (NB). Experimental results show that the improved SSO algorithm performs better than SSO, social spider algorithm, and CMA‐ES algorithms. Additionally, the performance of the proposed hybrid classification method is better than those of NB and DT. [ABSTRACT FROM AUTHOR]

Summary: A secure decision tree twin support vector machine (DT‐TSVM) multi‐classification algorithm has been proposed in this paper for improving the reliability and security of the collected IoT data from multiple data providers. The multiclass secure DT‐TSVM algorithm has been employed to train a machine learning model using the encrypted training dataset. The training dataset is collected via a blockchain platform. A blockchain method has been adopted to construct a secure and reliable distributed platform among dataset providers. The Paillier homomorphic cryptosystem has been applied for encrypting the IoT dataset. Then, the dataset has been recorded on the distributed ledger. The secure DT‐TSVM algorithm's‐based train model effectiveness has been compared with the other two available algorithms, namely the multiclass binary support vector machine (MBSVM) and one‐to‐one SVM algorithms. The experiment results showed that the privacy‐preserving multiclass secure DT‐TSVM‐based model did not reduce the accuracy, but it increased the average precision and recall by 0.53% and 0.44% than MBSVM and 0.82% and 0.71% than one‐to‐one SVM, respectively. Further, the time consumption of data providers and data analysts did not change significantly with the increase of number of data provider. [ABSTRACT FROM AUTHOR]

Knowledge discovery from temporal data (e.g., time series) is among the most challenging problems in data mining. Compared to static representations like rules or decision trees, the temporal component greatly increases the pattern diversity. It is important to keep the human perception of time flow in mind when representing temporal patterns, otherwise we open the floodgates to misinterpretation and misconception. This article gives an overview of temporal interval patterns, which are considered as being a well-suited mechanism of knowledge representation, and focusses on the various pattern representation languages. Four typical phenomena in temporal data, and how the pattern languages can cope with them, are discussed. Given the domain knowledge, this provides the reader some guidance on which pattern language may be best-suited for a given application. WIREs Data Mining Knowl Discov 2014, 4:196-212. doi: 10.1002/widm.1122 Conflict of interest: The authors have declared no conflicts of interest for this article. For further resources related to this article, please visit the . [ABSTRACT FROM AUTHOR]

In many data mining tasks, the goal is to classify entities into a set of pre-defined groups (classes). A second and equally important goal is the interpretation, i.e. understanding the nature of the population aggregated in each class. These tasks are rendered even more complex when there is no a-priori information regarding the right classification. The current paper is based on two concepts: (1) Bounded-Rationality theory which implements an S-shaped function that represents human logic as a saliency measure to determine the substantial features that characterize each potential group and (2) Classification by clustering (CBC) that applies Decision Tree-like classification in unsupervised clustering problems, where neither an a-priori classification nor target-attributes are known in advance. In the context of these two concepts, the current research contributes: (1) by expanding the saliency measure to all possible types of variables (nominal as well as numerical), (2) by evaluating, using five datasets, a composite model that combines the CBC method and the saliency concept. The findings show that by using clustering algorithms for classification tasks (CBC method) the results are as accurate as those obtained by conventional Decision Trees, but with a better saliency factor. [ABSTRACT FROM AUTHOR]

Dissolved gas analysis is an important way to diagnose transformer faults. Compared with the method of establishing a single classifier based on artificial intelligence for diagnosis, ensemble learning (EL) can combine multiple classifiers to achieve stronger generalization ability and better diagnostic performance. But the traditional EL belongs to homogenous ensemble in which the base learners are based on the same algorithm, so this kind of EL method lacks the differences among the base learners, as well as systematic combination strategy. For this problem, in the paper the Stacking ensemble strategy is applied to fault diagnosis. Multilayer perceptron, k‐nearest neighbor, decision tree and support vector machine are used as component learners, and random forest algorithm is used as a combination strategy to establish a Stacking diagnosis model. In addition, homogenous ensemble methods are applied to the above four algorithms. In the method, the content of five characteristic gases are taken as the input characteristic parameters. Primary diagnostic results can be obtained with each base classifier. Then the meta‐learner random forest model organizes the base classifiers, and uses the primary diagnostic output as the input of the meta‐learner for secondary diagnosis to get the final diagnosis. The experimental results show that the ensemble of multiple heterogeneous component learners can enhance the generalization ability of the model, and the diagnostic accuracy is better than single classifier and the homogenous ensemble classifier. © 2020 Institute of Electrical Engineers of Japan. Published by Wiley Periodicals LLC. [ABSTRACT FROM AUTHOR]

Copyright of Conservation Biology is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Target specificity and generally good tolerability of therapeutic proteins (TPs) present desirable treatment opportunities for pediatric patients. However, little is known on the ontogeny of processes related to the pharmacokinetics (PK) and disposition of TPs. The science, regulatory requirements and strategy of developing TPs for children are evolving. Our current review of TPs, (with focus on monoclonal antibodies and fusion proteins) that were approved for pediatric use indicates that dose-selection for pediatric pivotal studies is often based on adult PK information alone. This approach might not be sufficient if more complex PK properties than simple linear PK are present. Body weight-based dosing for pediatric patients directly scaled down from adult dosing can lead to under-exposure in young pediatric patients who are usually in the lowest body-weight range. Tiered-fixed dosing can be reasonably effective for TPs in achieving comparable exposure in children over a wide age range. The uniqueness of the pediatric population, the practical challenges in conducting clinical studies in this population, as well as regulations from health authorities warrant including pharmacometrics as an integral component of pediatric drug development. We propose a framework distinct from previous proposals, to guide clinical pharmacology strategy for pediatric drug development specifically for TPs. [ABSTRACT FROM AUTHOR]

A decision tree is a classification algorithm that automatically derives a hierarchy of partition rules with respect to a target attribute of a large dataset. However, spatial autocorrelation makes conventional decision trees underperform for geographical datasets as the spatial distribution is not taken into account. The research presented in this paper introduces the concept of a spatial decision tree based on a spatial diversity coefficient that measures the spatial entropy of a geo-referenced dataset. The principle of this solution is to take into account the spatial autocorrelation phenomena in the classification process, within a notion of spatial entropy that extends the conventional notion of entropy. Such a spatial entropy-based decision tree integrates the spatial autocorrelation component and generates a classification process adapted to geographical data. A case study oriented to the classification of an agriculture dataset in China illustrates the potential of the proposed approach. [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]

In the supervised classification, large training data are very common, and decision trees are widely used. However, as some bottlenecks such as memory restrictions, time complexity, or data complexity, many supervised classifiers including classical C4.5 tree cannot directly handle big data. One solution for this problem is to design a highly parallelized learning algorithm. Motivated by this, we propose a parallelized C4.5 decision tree algorithm based on MapReduce (MR-C4.5-Tree) with 2 parallelized methods to build the tree nodes. First, an information entropy-based parallelized attribute selection method (MR-A-S) on several subsets for MR-C4.5-Tree is proposed to confirm the best splitting attribute and the cut points. Then, a data splitting method (MR-D-S) in parallel is presented to partition the training data into subsets. At last, we introduce the MR-C4.5-Tree learning algorithm that grows in a top-down recursive way. Besides, the depth of the constructed decision tree, the number of samples and the maximal class probability in each tree node are used as the termination conditions to avoid the over-partitioning problem. Experimental studies show the feasibility and the good performance of the proposed parallelized MR-C4.5-Tree algorithm. [ABSTRACT FROM AUTHOR]