Random Undersampling Research Articles

Investigating the Effect of Class Balancing Methods on the Performance of Machine Learning Techniques: Credit Risk Application

Credit risk arises as a result of the failure of the loans given by banks to the customers to fulfill their obligations at the end of the specified term. Technological advances allow the use of machine learning methods in various sectors. These methods aim to facilitate the identification of customers at risk with the system adapted to the creditworthiness processes of banks. For this purpose, in order to make the most appropriate evaluation in the lending process of banks, re-sampling techniques to eliminate the problem of class imbalance encountered in unbalanced data sets were made balanced and their effects on machine learning were investigated. During the implementation phase, German, Australian and HMEQ credit data sets were used. Different machine learning classification methods such as Logistic Regression (LR), K-Narest Neighbor (KNN), Naive Bayes (NB), Support Vector Machines (SVM), Multilayer Perceptron (MLP), Decision Trees (DT), Random Forests (RF), Gradient Boosting Decision Trees (GBDT), Extremely Randomized Trees, Hard and Soft Voting were used to detect risky customers. The problem of class imbalance was balanced with resampling and hybrid techniques such as Random Oversampling (ROS), Random Undersampling (RUS), Balanced Bagging Classifier (BBC), SMOTE-Tomek Links and SMOTE-ENN. In this context, the performances of three different data sets were examined in four different scenarios. As a result of the study, the hybrid method, in which oversampling and undersampling methods are used together for the class balancing problem, showed the best classification performance among machine learning techniques.

An investigation using resampling techniques and explainable machine learning to minimize fire losses in residential buildings

Urban residential fires seriously threaten public safety, causing significant property damage and severely impacting urban sustainability. To enhance the understanding of urban residential fire risks, a framework that combines tree-based machine learning algorithms and resampling techniques is proposed to predict damage and casualties in residential building fires. All algorithms achieved similar results on the original dataset, with 86 % and 93 % accuracy and the highest average F1 scores of 61 % and 51 %, respectively. Various resampling techniques addressed the issue of data imbalance, with the combination of random undersampling and SMOTE achieving the best model performance, elevating the average F1 scores to 75 % and 77 %, representing improvements of 14 % and 26 % over the original dataset, respectively. Furthermore, the internal mechanisms of the model were explored using the explainable Shapley additive explanations, which identified the key features influencing model outputs. Additionally, the study revealed significant heterogeneity in different regions, sources of ignition, causes of fires, types of residences, locations of fire origin, and types of households. This research not only improves emergency response strategies for urban residential fires but also provides tailored fire safety policies to reduce risks in various urban environments effectively.

Attenuating majority attack class bias using hybrid deep learning based IDS framework

In real-time application domains, like finance, healthcare and defence, delay in service or stealing information may lead to unrecoverable consequences. So, early detection of intrusion is important to prevent security breaches. In recent days, anomaly-based intrusion detection using Hybrid Deep Learning approaches are becoming more popular. The most used benchmark datasets in the literature are NSL-KDD and UNSW-NB15, and these datasets are imbalanced. The models built on imbalanced datasets may lead to biased results towards majority classes by neglecting the minority class, even though they are equally important. In many cases, high accuracy is achieved for majority classes in the imbalanced datasets. But, the class-level performances are poor with respect to the minority class. The class balancing will also play an important role in attenuating the bias in prediction for imbalanced datasets. In this paper, a Hybrid Deep Learning Based Intrusion Detection (HDLBID) framework is proposed with CNN-BiLSTM combination. The four techniques, namely, Random Oversampling (ROS), ADASYN, SMOTE, and SMOTE-Tomek, are used for class balancing in the proposed HDLBID framework. The proposed HDLBID with SMOTE-Tomek achieves an overall accuracy of 99.6% with NSL-KDD and 89.02% for UNSW-NB15. It results in an improvement of 13.67% for NSL-KDD and 10.62% for UNSW-NB15 over the existing recent related models. In the proposed HDLBID, in addition to overall accuracy, the class-level F1 score is also calculated. A comparative study is presented to show the effectiveness of balancing dataset compared to imbalanced dataset, and observed that the SMOTE-Tomek class balancing comparatively performed well. An improvement of 37.43% is observed in the U2R class of the NSL-KDD dataset and 61.65% improvement is seen in the Worms class of the UNSW-NB15 dataset, both with SMOTE-Tomek class balancing. Therefore, the proposed HDLBID with SMOTE-Tomek class balancing reports the best results in terms of overall accuracy compared to existing recent related approaches. Also, in terms of class-level analysis, HDLBID reports best results with SMOTE-Tomek over imbalanced version of datasets.

Leveraging multi-site electronic health data for characterization of subtypes: a pilot study of dementia in the N3C Clinical Tenant.

To provide a foundational methodology for differentiating comorbidity patterns in subphenotypes through investigation of a multi-site dementia patient dataset. Employing the National Clinical Cohort Collaborative Tenant Pilot (N3C Clinical) dataset, our approach integrates machine learning algorithms-logistic regression and eXtreme Gradient Boosting (XGBoost)-with a diagnostic hierarchical model for nuanced classification of dementia subtypes based on comorbidities and gender. The methodology is enhanced by multi-site EHR data, implementing a hybrid sampling strategy combining 65% Synthetic Minority Over-sampling Technique (SMOTE), 35% Random Under-Sampling (RUS), and Tomek Links for class imbalance. The hierarchical model further refines the analysis, allowing for layered understanding of disease patterns. The study identified significant comorbidity patterns associated with diagnosis of Alzheimer's, Vascular, and Lewy Body dementia subtypes. The classification models achieved accuracies up to 69% for Alzheimer's/Vascular dementia and highlighted challenges in distinguishing Dementia with Lewy Bodies. The hierarchical model elucidates the complexity of diagnosing Dementia with Lewy Bodies and reveals the potential impact of regional clinical practices on dementia classification. Our methodology underscores the importance of leveraging multi-site datasets and tailored sampling techniques for dementia research. This framework holds promise for extending to other disease subtypes, offering a pathway to more nuanced and generalizable insights into dementia and its complex interplay with comorbid conditions. This study underscores the critical role of multi-site data analyzes in understanding the relationship between comorbidities and disease subtypes. By utilizing diverse healthcare data, we emphasize the need to consider site-specific differences in clinical practices and patient demographics. Despite challenges like class imbalance and variability in EHR data, our findings highlight the essential contribution of multi-site data to developing accurate and generalizable models for disease classification.

Addressing Class Imbalance in Customer Response Modeling Using Random and Clustering-Based Undersampling and SVM

Addressing Class Imbalance in Customer Response Modeling Using Random and Clustering-Based Undersampling and SVM

A Fault Prediction Method for Secondary Electric Power Equipment Based on Multi-Round Undersampling Random Forest

Abstract It is necessary to predict the fault states of the large number of power secondary equipment using intelligent methods. However, when compared to large-scale equipment, the number of negative samples representing faulty equipment is significantly smaller than the number of positive samples representing normal equipment. This leads to a pronounced imbalance between positive and negative samples in the task of fault prediction. In this paper, we propose a multi-round undersampling random forest method to predict the fault situations of secondary electric power equipment. First, we collect data from historical power system logs to build the dataset for power secondary equipment and preprocess it. The undersampling method is utilized to generate a balanced dataset of secondary power equipment with a smaller sample size. We generate multiple balanced datasets through rounds of random undersampling without replacement to train multiple random forest models. Subsequently, we predict fault situations in secondary electric power equipment through comprehensive decision-making by these multiple random forest models. We employ a real dataset from the power system of Chongqing, China, for experimental validation. The results demonstrate the superiority of our method over other machine learning prediction models used for comparison.

Evaluation of Sampling Algorithms Used for Bayesian Uncertainty Quantification of Molecular Dynamics Force Fields.

New Bayesian parameter estimation methods have the capability to enable more physically realistic and reliable molecular dynamics (MD) simulations by providing accurate estimates of uncertainties of force-field (FF) parameters and associated properties. However, the choice of which Bayesian parameter estimation algorithm to use has not been widely investigated, despite its impact on the effective exploration of parameter space. Here, using a case example of the Embedded Atom Method (EAM) FF parameters, we investigated the ramifications of several of the algorithm choices. We found that Ensemble Slice Sampling (ESS) and Affine-Invariant Ensemble Sampling (AIES) demonstrate a new level of superior performance, culminating in more accurate parameter and property estimations with tighter uncertainty bounds, compared to traditional methods such as Metropolis-Hastings (MH), Gradient Search (GS), and Uniform Random Sampler (URS). We demonstrate that Bayesian Uncertainty Quantification with ESS and AIES leads to significantly more accurate and reliable predictions of the FF parameters and properties. The results suggest that ESS and AIES should be used to obtain more accurate parameter and uncertainty estimations while providing deeper physical insights.

Penerapan Random Oversampling dan Algoritma Boosting untuk Memprediksi Kualitas Buah Jeruk

According to the 2019 data, global orange production has increased significantly, reaching 79 million tons. However, despite the availability of various types of oranges in Indonesian markets, many vendors still sell low-quality oranges. To address this issue, researchers have applied random oversampling and boosting algorithms to predict orange quality, using the public Orange Quality Analysis Dataset. This study uses random oversampling to address data imbalance and combines it with boosting algorithms like Adaboost, Gradient Boosting, Light GBM, and CatBoost. The data features considered include size, weight, sweetness level, acidity level, and others. The accuracy of the boosting algorithms used varied, with CatBoost showing the highest accuracy rate of 91.42%. The hope is that this research can help orange producers create high-quality products and reduce the occurrence of low-quality oranges, ultimately providing consumers with better oranges. Additionally, this can help producers market their oranges both domestically and internationally.

Classification of autonomous vehicle crash severity: Solving the problems of imbalanced datasets and small sample size

Only a few researchers have shown how environmental factors and road features relate to Autonomous Vehicle (AV) crash severity levels, and none have focused on the data limitation problems, such as small sample sizes, imbalanced datasets, and high dimensional features. To address these problems, we analyzed an AV crash dataset (2019 to 2021) from the California Department of Motor Vehicles (CA DMV), which included 266 collision reports (51 of those causing injuries). We included external environmental variables by collecting various points of interest (POIs) and roadway features from Open Street Map (OSM) and Data San Francisco (SF). Random Over-Sampling Examples (ROSE) and the Synthetic Minority Over-Sampling Technique (SMOTE) methods were used to balance the dataset and increase the sample size. These two balancing methods were used to expand the dataset and solve the small sample size problem simultaneously. Mutual information, random forest, and XGboost were utilized to address the high dimensional feature and the selection problem caused by including a variety of types of POIs as predictive variables. Because existing studies do not use consistent procedures, we compared the effectiveness of using the feature-selection preprocessing method as the first process to employing the data-balance technique as the first process.Our results showed that AV crash severity levels are related to vehicle manufacturers, vehicle damage level, collision type, vehicle movement, the parties involved in the crash, speed limit, and some types of POIs (areas near transportation, entertainment venues, public places, schools, and medical facilities). Both resampling methods and three data preprocessing methods improved model performance, and the model that used SMOTE and data-balancing first was the best. The results suggest that over-sampling and the feature selection method can improve model prediction performance and define new factors related to AV crash severity levels.

A Hybrid Model for Detecting Intrusions on Network Logs

The presence of malicious traffic presents a substantial risk to network systems and the integrity of confidential information. Organisations may enhance their protection against threats and mitigate the possible impact of malicious traffic on their networks by maintaining vigilance, deploying comprehensive security measures, and cultivating a cybersecurity-aware culture. The purpose of this study is to propose a theoretical framework for identifying and analysing potentially harmful network traffic within a network system. In order to identify and classify various types of malicious network traffic in a multi-class setting, we employed a dataset consisting of nine distinct categories of network system attacks. In order to optimise the performance of the model, an exploratory data analysis is conducted on the dataset. Exploratory data analysis (EDA) was employed to assess various aspects like the presence of missing values, correlation among characteristics, data imbalance, and identification of significant features. The findings derived from the exploratory data analysis indicate that the dataset exhibits an imbalance, which, if left unaddressed, may result in overfitting. The data imbalance was addressed with the implementation of the RandomOverSampling approach in Python, which involved executing random oversampling. Following the resolution of the data imbalance, a random forest classifier was employed to extract significant features from the dataset. In this study, a total of ten characteristics were extracted based on the ranking provided by the random forest model. The features that were extracted were utilised in the training process of the suggested model, which aims to identify and detect malicious activity within a network system. The findings of the model indicate a much improved level of accuracy in identifying malicious traffic within a network system, with an accuracy rate of 99.99%. Furthermore, the precision, recall, and F1-score metrics also demonstrate a consistent accuracy rate of 99.99%.

Understanding random resampling techniques for class imbalance correction and their consequences on calibration and discrimination of clinical risk prediction models

ObjectiveClass imbalance is sometimes considered a problem when developing clinical prediction models and assessing their performance. To address it, correction strategies involving manipulations of the training dataset, such as random undersampling or oversampling, are frequently used. The aim of this article is to illustrate the consequences of these class imbalance correction strategies on clinical prediction models’ internal validity in terms of calibration and discrimination performances. MethodsWe used both heuristic intuition and formal mathematical reasoning to characterize the relations between conditional probabilities of interest and probabilities targeted when using random undersampling or oversampling. We propose a plug-in estimator that represents a natural correction for predictions obtained from models that have been trained on artificially balanced datasets (“naïve” models). We conducted a Monte Carlo simulation with two different data generation processes and present a real-world example using data from the International Stroke Trial database to empirically demonstrate the consequences of applying random resampling techniques for class imbalance correction on calibration and discrimination (in terms of Area Under the ROC, AUC) for logistic regression and tree-based prediction models. ResultsAcross our simulations and in the real-world example, calibration of the naïve models was very poor. The models using the plug-in estimator generally outperformed the models relying on class imbalance correction in terms of calibration while achieving the same discrimination performance. ConclusionRandom resampling techniques for class imbalance correction do not generally improve discrimination performance (i.e., AUC), and their use is hard to justify when aiming at providing calibrated predictions. Improper use of such class imbalance correction techniques can lead to suboptimal data usage and less valid risk prediction models.

Improving accuracy of code smells detection using machine learning with data balancing techniques

Code smells indicate potential symptoms or problems in software due to inefficient design or incomplete implementation. These problems can affect software quality in the long-term. Code smell detection is fundamental to improving software quality and maintainability, reducing software failure risk, and helping to refactor the code. Previous works have applied several prediction methods for code smell detection. However, many of them show that machine learning (ML) and deep learning (DL) techniques are not always suitable for code smell detection due to the problem of imbalanced data. So, data imbalance is the main challenge for ML and DL techniques in detecting code smells. To overcome these challenges, this study aims to present a method for detecting code smell based on DL algorithms (Bidirectional Long Short-Term Memory (Bi-LSTM) and Gated Recurrent Unit (GRU)) combined with data balancing techniques (random oversampling and Tomek links) to mitigate data imbalance issue. To establish the effectiveness of the proposed models, the experiments were conducted on four code smells datasets (God class, data Class, feature envy, and long method) extracted from 74 open-source systems. We compare and evaluate the performance of the models according to seven different performance measures accuracy, precision, recall, f-measure, Matthew’s correlation coefficient (MCC), the area under a receiver operating characteristic curve (AUC), the area under the precision–recall curve (AUCPR) and mean square error (MSE). After comparing the results obtained by the proposed models on the original and balanced data sets, we found out that the best accuracy of 98% was obtained for the Long method by using both models (Bi-LSTM and GRU) on the original datasets, the best accuracy of 100% was obtained for the long method by using both models (Bi-LSTM and GRU) on the balanced datasets (using random oversampling), and the best accuracy 99% was obtained for the long method by using Bi-LSTM model and 99% was obtained for the data class and Feature envy by using GRU model on the balanced datasets (using Tomek links). The results indicate that the use of data balancing techniques had a positive effect on the predictive accuracy of the models presented. The results show that the proposed models can detect the code smells more accurately and effectively.

Dealing imbalance dataset problem in sentiment analysis of recession in Indonesia

<span lang="EN-US">Global recession news dominates social media, particularly in Indonesia, with social news platforms on Twitter generating public responses and <br /> re-tweetings on the issue. Mining these opinions from Twitter using a sentiment analysis approach yields invaluable insights. The research stages included data collection, pre-processing, data labeling using the <br /> lexical-based method like valence aware dictionary and sentiment reasoner (VADER) and TextBlob, sampling techniques using synthetic minority oversampling technique (SMOTE) and random over sampling (ROS) before and after splitting data, and modeling using machine learning such as support vector machines (SVM), k-nearest neighbour (KNN), naive Bayes, and model evaluation. The problem is that almost 300,000 data collected from NodeXL are unbalanced. The findings show that models with balanced datasets show better model evaluation results. The sampling technique was carried out before and after splitting the data. The model evaluation results show that the Bernoulli-naive Bayes algorithm, with the VADER labeling technique, and the SMOTE sampling technique after splitting data, obtains the best accuracy of 84%, and using the ROS technique obtains an accuracy of 81%. On the other hand, with the SMOTE and ROS technique before splitting data on the SVM algorithm, it gets the best accuracy of 93% from before if only using SVM only reached 84%.</span>

Optimizing the Performance of AI Model for Non-Invasive Continuous Glucose Monitoring: Hyperparameter Tuning and Random Oversampling Approach

Diabetes Mellitus (DM) as a non-communicable disease (NCD) continues to increase every year. Continuous glucose monitoring (CGM) is essential for effective DM management. However, existing disposable glucose monitoring methods still rely on invasive techniques, cause pain, and lack continuous monitoring capabilities. On the other hand, non-invasive techniques are not feasible for CGM due to the biometric data's complexity and the classification system's inadequate performance. This study aims to develop a non-invasive technology to improve the performance of a non-invasive blood glucose classification system using Artificial Intelligence (AI), specifically Convolutional Neural Network (CNN) and an oversampling technique. The oversampling technique could improve data quantity by balancing the amount of data for each class. This study recruited twenty-three participants in the age range of 20 to 22 years comprising seven females and fifteen males. During data recording sessions, blood glucose levels were simultaneously assessed using a gold-standard glucometer and a non-invasive CGM prototype. The proposed CNN model successfully improved the classification accuracy of non-invasive blood glucose monitoring significantly. With the implementation of oversampling for augmenting the data, the accuracy of the proposed model increased to more than 88%. This study concludes that non-invasive approaches combined with AI technology have the potential to provide a convenient and pain-free alternative to traditional monitoring methods, significantly improving diabetes management and enhancing the overall quality of life for those affected by this condition. These findings could revolutionize the field of diabetes management, offering a more comfortable and accurate monitoring solution that could potentially transform the lives of millions of diabetes patients.

Smart Data Driven Decision Trees Ensemble Methodology for Imbalanced Big Data

Differences in data size per class, also known as imbalanced data distribution, have become a common problem affecting data quality. Big Data scenarios pose a new challenge to traditional imbalanced classification algorithms, since they are not prepared to work with such amount of data. Split data strategies and lack of data in the minority class due to the use of MapReduce paradigm have posed new challenges for tackling the imbalance between classes in Big Data scenarios. Ensembles have been shown to be able to successfully address imbalanced data problems. Smart Data refers to data of enough quality to achieve high-performance models. The combination of ensembles and Smart Data, achieved through Big Data preprocessing, should be a great synergy. In this paper, we propose a novel Smart Data driven Decision Trees Ensemble methodology for addressing the imbalanced classification problem in Big Data domains, namely SD_DeTE methodology. This methodology is based on the learning of different decision trees using distributed quality data for the ensemble process. This quality data is achieved by fusing random discretization, principal components analysis, and clustering-based random oversampling for obtaining different Smart Data versions of the original data. Experiments carried out in 21 binary adapted datasets have shown that our methodology outperforms random forest.

3-Way hybrid analysis using clinical and magnetic resonance imaging for early diagnosis of Alzheimer’s disease

Alzheimer’s is a progressive neurodegenerative disorder that leads to cognitive impairment and ultimately death. To select the most effective treatment options, it is crucial to diagnose and classify the disease early, as current treatments can only delay its progression. However, previous research on Alzheimer’s disease (AD) has had limitations, such as inaccuracies and reliance on a small, unbalanced binary dataset. In this study, we aimed to evaluate the early stages of AD using three multiclass datasets: OASIS, EEG, and ADNI MRI. The research consisted of three phases: pre-processing, feature extraction, and classification using hybrid learning techniques. For the OASIS and ADNI MRI datasets, we computed the mean RGB value and used an averaging filter to enhance the images. We balanced and augmented the dataset to increase its size. In the case of the EEG dataset, we applied a band-pass filter for digital filtering to reduce noise and also balanced the dataset using random oversampling. To extract and classify features, we utilized a hybrid technique consisting of four algorithms: AlexNet-MLP, AlexNet-ETC, AlexNet-AdaBoost, and AlexNet-NB. The results showed that the AlexNet-ETC hybrid algorithm achieved the highest accuracy rate of 95.32% for the OASIS dataset. In the case of the EEG dataset, the AlexNet-MLP hybrid algorithm outperformed other approaches with the highest accuracy of 97.71%. For the ADNI MRI dataset, the AlexNet-MLP hybrid algorithm achieved an accuracy rate of 92.59%. Comparing these results with the current state of the art demonstrates the effectiveness of our findings.

Detecting contract cheating through linguistic fingerprint

Contract cheating, the act of students enlisting others to complete academic assignments on their behalf, poses a significant challenge in academic settings, undermining the integrity of education and assessment. It involves submitting work that is falsely represented as the student’s own, thus violating academic standards and ethics. The advent of artificial intelligence-based language models, such as ChatGPT, has raised concerns about the potential impact of contract cheating. As these language models can generate human-like text with ease, there are concerns about their role in facilitating and increasing contract cheating incidents. Innovative approaches are thus needed to detect contract cheating and address its implications for academic integrity. This study introduces a machine learning (ML) model focused on identifying deviations from a learner’s unique writing style (or their linguistic fingerprint) to detect contract cheating, complementing traditional plagiarism detection methods. The study involved 150 learners majoring in engineering and business who were studying English as a foreign language at a college in Saudi Arabia. The participants were asked to produce descriptive essays in English within a consistent genre over one semester. The proposed approach involved data preprocessing, followed by transformation using Term Frequency-Inverse Document Frequency (TF-IDF). To address data imbalance, random oversampling was applied, and logistic regression (LR) was trained with optimal hyperparameters obtained through grid search. Performance evaluation was conducted using various metrics. The results showed that the ML model was effective in identifying non-consistent essays with improved accuracy after implementing random oversampling. The LR model achieved an accuracy of 98.03%, precision of 98.52%, recall of 98.03%, and F1-score of 98.24%. The proposed ML model shows promise as an indicator of contract cheating incidents, providing an additional tool for educators and institutions to uphold academic integrity. However, it is essential to interpret the model results cautiously, as they do not constitute unequivocal evidence of cheating but rather serve as grounds for further investigation. We also emphasize the ethical implications of such approaches and suggest avenues for future research to explore the model’s applicability among first-language writers and to conduct longitudinal studies on second-language learners’ language development over longer periods.

Rice phenology monitoring via ensemble classification for an extremely imbalanced multiclass dataset of hybrid remote sensing

This research provides a new approach for monitoring crops, especially rice planting, using hybrid remote sensing by aligning direct observations via Area Sampling Frame (ASF) surveys and earth observations via the LANDSAT 8 spectral indices. ASF surveys classify the land cover into eight groups, four associated with rice phenology. There's probably an imbalance between the classes in the multi-class ASF dataset. Often, when dealing with such extremely multiclass datasets, the performance of conventional classifiers is not adequate. Therefore, this study adopts a resampling technique to improve ensemble classification performance for handling extremely imbalanced multiclass ASF datasets. This research has added resampling methods, ROS (Random Over Sampling) and SCUT (SMOTE and Cluster-based Undersampling Technique), to the ensemble model, namely Bagging and Random Forest (RF), which have an impact on improving classification performance. The results showed that ROS-RF performed well in classifying multiclass on the ASF Survey with accuracy (90%), average sensitivity (90%), average specificity (99%), and balanced accuracy (94%). The variable importance (VI) with the highest ranking based on the Mean Decrease Gini (MDG) is the Modified Normalized Difference Water Index, which was recorded by the LANDSAT 8 satellite in one period before (MDWIt-1). However, the nine variables included in the classification generally contribute almost the same level of importance. The recommended classification strategy for highly imbalanced multi-class datasets is ROS-RF, which contributes to improving RF classification performance. The initial step was repeating minority samples until the class distribution was balanced before executing the classification stage.

Examining Factors Contributing to Motorcycle Collisions with Left-Turning Vehicles at Urban Intersection Locations

Motorcycle crashes account for a significant proportion of traffic-related fatalities on U.S. roadways. Compared with motor vehicles, motorcycles traveling straight ahead are more susceptible to collisions with left-turning vehicles at intersections (note – in a system where traffic travels on the right-hand side of the road). The limited knowledge of the causes and influences of this specific type of crash deters efforts to improve motorcycle safety and is partly influenced by two issues. First, significant variables are unknown; second, motorcycles comprise a small proportion of vehicles in the traffic stream. This study sought to understand the factors that may contribute to the disproportionate crash risk left-turning vehicles pose for motorcyclists while accounting for the imbalance of vehicle proportions. Data containing motorcycle-motor vehicle and motor vehicle-motor vehicle crashes involving left-turning motor vehicles at intersections in South Florida were collected from 2015 to 2017. The study applied logistic regression on a balanced dataset generated using the random oversampling technique. The proposed model improved the predictive accuracy and enabled the identification of factors contributing to motorcycle crashes with left-turning vehicles. A Bayesian network analysis was also applied to the balanced data to analyze the interrelationship of factors associated with motorcycle crashes with left-turning vehicles. Results indicated that the type of intersection and traffic control, time of day, age of drivers, sex of the motorcyclist, roadway type, and weather were significantly associated with motorcyclists’ susceptibility to collisions with left-turning vehicles. Recognizing these attributes could help devise engineering measures and policies for promoting motorcycle safety.

Improving Credit Card Fraud Detection with Data Reduction Approaches

Detecting fraudulent activities in credit card transactions can be challenging due to issues like high dimensionality and class imbalance that are often present in the datasets. To address these challenges, data reduction techniques such as data sampling and feature selection have become essential. In this study, we compare four approaches for data reduction: using data sampling alone, employing feature selection alone, applying data sampling followed by feature selection, and using feature selection followed by data sampling. Additionally, we include results using all features. We build classification models using five Decision Tree-based classifiers and Logistic Regression, and evaluate their performance using two performance metrics: the Area Under the Receiver Operating Characteristic Curve (AUC) and the Area under the Precision–Recall Curve (AUPRC). In this work, we adopt ensemble supervised feature selection (SFS) techniques and Random Undersampling (RUS) for data reduction. The experimental results demonstrate that all four data reduction techniques have the potential to improve the performance of classifiers. These results are valuable since the classifiers available are dependent upon application domains, computing environments, and licensing agreements. However, these techniques can be applied independently of all these dependencies. We recommend utilizing the ensemble SFS followed by RUS (SFS–RUS) approach as the preferred data reduction method due to its ability to run feature selection and data sampling in parallel. Additionally, we find that XGBoost and CatBoost outperform other classifiers.