scholarly journals RDPVR: Random Data Partitioning with Voting Rule for Machine Learning from Class-Imbalanced Datasets

Electronics ◽  
2022 ◽  
Vol 11 (2) ◽  
pp. 228
Author(s):  
Ahmad B. Hassanat ◽  
Ahmad S. Tarawneh ◽  
Samer Subhi Abed ◽  
Ghada Awad Altarawneh ◽  
Malek Alrashidi ◽  
...  

Since most classifiers are biased toward the dominant class, class imbalance is a challenging problem in machine learning. The most popular approaches to solving this problem include oversampling minority examples and undersampling majority examples. Oversampling may increase the probability of overfitting, whereas undersampling eliminates examples that may be crucial to the learning process. We present a linear time resampling method based on random data partitioning and a majority voting rule to address both concerns, where an imbalanced dataset is partitioned into a number of small subdatasets, each of which must be class balanced. After that, a specific classifier is trained for each subdataset, and the final classification result is established by applying the majority voting rule to the results of all of the trained models. We compared the performance of the proposed method to some of the most well-known oversampling and undersampling methods, employing a range of classifiers, on 33 benchmark machine learning class-imbalanced datasets. The classification results produced by the classifiers employed on the generated data by the proposed method were comparable to most of the resampling methods tested, with the exception of SMOTEFUNA, which is an oversampling method that increases the probability of overfitting. The proposed method produced results that were comparable to the Easy Ensemble (EE) undersampling method. As a result, for solving the challenge of machine learning from class-imbalanced datasets, we advocate using either EE or our method.

2016 ◽  
Vol 7 (2) ◽  
pp. 43-71 ◽  
Author(s):  
Sangeeta Lal ◽  
Neetu Sardana ◽  
Ashish Sureka

Logging is an important yet tough decision for OSS developers. Machine-learning models are useful in improving several steps of OSS development, including logging. Several recent studies propose machine-learning models to predict logged code construct. The prediction performances of these models are limited due to the class-imbalance problem since the number of logged code constructs is small as compared to non-logged code constructs. No previous study analyzes the class-imbalance problem for logged code construct prediction. The authors first analyze the performances of J48, RF, and SVM classifiers for catch-blocks and if-blocks logged code constructs prediction on imbalanced datasets. Second, the authors propose LogIm, an ensemble and threshold-based machine-learning model. Third, the authors evaluate the performance of LogIm on three open-source projects. On average, LogIm model improves the performance of baseline classifiers, J48, RF, and SVM, by 7.38%, 9.24%, and 4.6% for catch-blocks, and 12.11%, 14.95%, and 19.13% for if-blocks logging prediction.


10.2196/15601 ◽  
2019 ◽  
Vol 7 (4) ◽  
pp. e15601 ◽  
Author(s):  
Quazi Abidur Rahman ◽  
Tahir Janmohamed ◽  
Hance Clarke ◽  
Paul Ritvo ◽  
Jane Heffernan ◽  
...  

Background Pain volatility is an important factor in chronic pain experience and adaptation. Previously, we employed machine-learning methods to define and predict pain volatility levels from users of the Manage My Pain app. Reducing the number of features is important to help increase interpretability of such prediction models. Prediction results also need to be consolidated from multiple random subsamples to address the class imbalance issue. Objective This study aimed to: (1) increase the interpretability of previously developed pain volatility models by identifying the most important features that distinguish high from low volatility users; and (2) consolidate prediction results from models derived from multiple random subsamples while addressing the class imbalance issue. Methods A total of 132 features were extracted from the first month of app use to develop machine learning–based models for predicting pain volatility at the sixth month of app use. Three feature selection methods were applied to identify features that were significantly better predictors than other members of the large features set used for developing the prediction models: (1) Gini impurity criterion; (2) information gain criterion; and (3) Boruta. We then combined the three groups of important features determined by these algorithms to produce the final list of important features. Three machine learning methods were then employed to conduct prediction experiments using the selected important features: (1) logistic regression with ridge estimators; (2) logistic regression with least absolute shrinkage and selection operator; and (3) random forests. Multiple random under-sampling of the majority class was conducted to address class imbalance in the dataset. Subsequently, a majority voting approach was employed to consolidate prediction results from these multiple subsamples. The total number of users included in this study was 879, with a total number of 391,255 pain records. Results A threshold of 1.6 was established using clustering methods to differentiate between 2 classes: low volatility (n=694) and high volatility (n=185). The overall prediction accuracy is approximately 70% for both random forests and logistic regression models when using 132 features. Overall, 9 important features were identified using 3 feature selection methods. Of these 9 features, 2 are from the app use category and the other 7 are related to pain statistics. After consolidating models that were developed using random subsamples by majority voting, logistic regression models performed equally well using 132 or 9 features. Random forests performed better than logistic regression methods in predicting the high volatility class. The consolidated accuracy of random forests does not drop significantly (601/879; 68.4% vs 618/879; 70.3%) when only 9 important features are included in the prediction model. Conclusions We employed feature selection methods to identify important features in predicting future pain volatility. To address class imbalance, we consolidated models that were developed using multiple random subsamples by majority voting. Reducing the number of features did not result in a significant decrease in the consolidated prediction accuracy.


Symmetry ◽  
2021 ◽  
Vol 13 (9) ◽  
pp. 1649
Author(s):  
Zhuang Li ◽  
Jingyan Qin ◽  
Xiaotong Zhang ◽  
Yadong Wan

Class imbalance, as a phenomenon of asymmetry, has an adverse effect on the performance of most machine learning and overlap is another important factor that affects the classification performance of machine learning algorithms. This paper deals with the two factors simultaneously, addressing the class overlap under imbalanced distribution. In this paper, a theoretical analysis is firstly conducted on the existing class overlap metrics. Then, an improved method and the corresponding metrics to evaluate the class overlap under imbalance distributions are proposed based on the theoretical analysis. A well-known collection of the imbalanced datasets is used to compare the performance of different metrics and the performance is evaluated based on the Pearson correlation coefficient and the ξ correlation coefficient. The experimental results demonstrate that the proposed class overlap metrics outperform other compared metrics for the imbalanced datasets and the Pearson correlation coefficient with the AUC metric of eight algorithms can be improved by 34.7488% in average.


2020 ◽  
Vol 8 (5) ◽  
pp. 1907-1916

Imbalanced data learning is a research area and day by day development is going on. Due to these researchers are motivated to pay attention to find efficient and adaptive methods for real-world problems. Machine learning, as well as data mining, is a field where researchers are finding different methods to solve problems related to imbalanced datasets and also the challenges faced in day to day life. The uneven class distribution in the dataset is the reason behind the degradation of performance in approaches used by data mining as well as machine learning. Continuous advancements of machine learning as well as mining data combining it with big data, a deep insight is required to understand the nature of learning imbalanced data. New challenges are emerging due to this development. Among the two approaches algorithm level and data level, the most popular approach compared to this is the hybrid approach. It is found that there is a bias for the majority class which affects the decision making task and overall accuracy of classification. The ensemble method is an efficient technique to deal with the uneven distribution of data. The aim of the paper is to presents the overview of class imbalance problems, solutions to handle it, open issues and challenges in learning imbalanced datasets. Based on the experiment conducted on one dataset it is found that ensemble technique along with other data-level methods gives good results. This hybrid method can be applied in many real-life applications like software defect prediction, behavior analysis, intrusion detection, medical diagnosis, etc. The paper further provides research directions in learning from the imbalanced dataset.


2021 ◽  
Vol 11 (2) ◽  
pp. 15-37
Author(s):  
Ankita Bansal ◽  
Makul Saini ◽  
Rakshit Singh ◽  
Jai Kumar Yadav

The tremendous amount of data generated through IoT can be imbalanced causing class imbalance problem (CIP). CIP is one of the major issues in machine learning where most of the samples belong to one of the classes, thus producing biased classifiers. The authors in this paper are working on four imbalanced datasets belonging to diverse domains. The objective of this study is to deal with CIP using oversampling techniques. One of the commonly used oversampling approaches is synthetic minority oversampling technique (SMOTE). In this paper, the authors have suggested modifications in SMOTE and proposed their own algorithm, SMOTE-modified (SMOTE-M). To provide a fair evaluation, it is compared with three oversampling approaches, SMOTE, adaptive synthetic oversampling (ADASYN), and SMOTE-Adaboost. To evaluate the performances of sampling approaches, models are constructed using four classifiers (K-nearest neighbour, decision tree, naive Bayes, logistic regression) on balanced and imbalanced datasets. The study shows that the results of SMOTE-M are comparable to that of ADASYN and SMOTE-Adaboost.


2017 ◽  
Vol 1 (1) ◽  
pp. 114
Author(s):  
M Bilal Shaikh ◽  
M Abdul Rehman ◽  
Attaullah Sahito

Distributed Machine Learning (DML) has gained its importance more than ever in this era of Big Data. There are a lot of challenges to scale machine learning techniques on distributed platforms. When it comes to scalability, improving the processor technology for high level computation of data is at its limit, however increasing machine nodes and distributing data along with computation looks as a viable solution. Different frameworks   and platforms are available to solve DML problems. These platforms provide automated random data distribution of datasets which miss the power of user defined intelligent data partitioning based on domain knowledge. We have conducted an empirical study which uses an EEG Data Set collected through P300 Speller component of an ERP (Event Related Potential) which is widely used in BCI problems; it helps in translating the intention of subject w h i l e performing any cognitive task. EEG data contains noise due to waves generated by other activities in the brain which contaminates true P300Speller. Use of Machine Learning techniques could help in detecting errors made by P300 Speller. We are solving this classification problem by partitioning data into different chunks and preparing distributed models using Elastic CV Classifier. To present a case of optimizing distributed machine learning, we propose an intelligent user defined data partitioning approach that could impact on the accuracy of distributed machine learners on average. Our results show better average AUC as compared to average AUC obtained after applying random data partitioning which gives no control to user over data partitioning. It improves the average accuracy of distributed learner due to the domain specific intelligent partitioning by the user. Our customized approach achieves 0.66 AUC on individual sessions and 0.75 AUC on mixed sessions, whereas random / uncontrolled data distribution records 0.63 AUC.


2020 ◽  
pp. 740-772
Author(s):  
Sangeeta Lal ◽  
Neetu Sardana ◽  
Ashish Sureka

Logging is an important yet tough decision for OSS developers. Machine-learning models are useful in improving several steps of OSS development, including logging. Several recent studies propose machine-learning models to predict logged code construct. The prediction performances of these models are limited due to the class-imbalance problem since the number of logged code constructs is small as compared to non-logged code constructs. No previous study analyzes the class-imbalance problem for logged code construct prediction. The authors first analyze the performances of J48, RF, and SVM classifiers for catch-blocks and if-blocks logged code constructs prediction on imbalanced datasets. Second, the authors propose LogIm, an ensemble and threshold-based machine-learning model. Third, the authors evaluate the performance of LogIm on three open-source projects. On average, LogIm model improves the performance of baseline classifiers, J48, RF, and SVM, by 7.38%, 9.24%, and 4.6% for catch-blocks, and 12.11%, 14.95%, and 19.13% for if-blocks logging prediction.


2019 ◽  
Vol 12 (10) ◽  
Author(s):  
Swati Narwane ◽  
Sudhir Sawarkar

Mathematics ◽  
2021 ◽  
Vol 9 (9) ◽  
pp. 936
Author(s):  
Jianli Shao ◽  
Xin Liu ◽  
Wenqing He

Imbalanced data exist in many classification problems. The classification of imbalanced data has remarkable challenges in machine learning. The support vector machine (SVM) and its variants are popularly used in machine learning among different classifiers thanks to their flexibility and interpretability. However, the performance of SVMs is impacted when the data are imbalanced, which is a typical data structure in the multi-category classification problem. In this paper, we employ the data-adaptive SVM with scaled kernel functions to classify instances for a multi-class population. We propose a multi-class data-dependent kernel function for the SVM by considering class imbalance and the spatial association among instances so that the classification accuracy is enhanced. Simulation studies demonstrate the superb performance of the proposed method, and a real multi-class prostate cancer image dataset is employed as an illustration. Not only does the proposed method outperform the competitor methods in terms of the commonly used accuracy measures such as the F-score and G-means, but also successfully detects more than 60% of instances from the rare class in the real data, while the competitors can only detect less than 20% of the rare class instances. The proposed method will benefit other scientific research fields, such as multiple region boundary detection.


Sign in / Sign up

Export Citation Format

Share Document