Balanced Class Research Articles

HANDLING HIGHLY IMBALANCED OUTPUT CLASS LABEL

In practice, a balanced target class is rare. However, an imbalanced target class can be handled by resampling the original dataset, either by oversampling/upsampling or undersampling/downsampling. A popular upsampling technique is Synthetic Minority Over-sampling Technique (SMOTE). This technique increases the minority class by generating synthetic class labels and assigned the class based on the K-Nearest Neighbour (K-NN). SMOTE upsampling can only upsample at most one minority class at a time, which means for a multiclass dataset, it needs to undergo multilayer SMOTE to balance the class label distribution. This paper aims to find a suitable method in handling imbalanced class using dataset from Fantasy Premier League (FPL) virtual player to predict price changes. The cleaned dataset has a highly imbalanced class distribution, where the frequency of “Price Remain Unchanged (PRU)” is higher than “Price Fall (PF)” and “Price Rise (PR)”. This paper compared between the baseline (original) dataset, SMOTE-applied dataset and shuffled, linear and stratified sampling in split train-test subset, based on a deep learning algorithm. This paper also proposed criteria of low values in standard deviation (distribution of true positive on each class label on accuracy) as a measurement for finding the best method in handling imbalanced class labels. As a result, multilayer SMOTE until all the classes distribution is the same, combined with stratified sampling in split training and testing subset, get the lower standard deviation (5.7873), high accuracy (80.06%) and less execution runtime (1 minute 41 seconds) compared to the original highly imbalanced dataset.

A quantum algorithm based on entanglement measure for classifying Boolean multivariate function into novel hidden classes

In this paper, we propose a novel algorithm that solves a generalized version of the Deutsch-Jozsa problem. The proposed algorithm has the potential to classify an oracle UF, that represents an unknown Boolean function on n Boolean variables, to one of 2n different classes instead of only two classes which are constant and balanced classes in the case of Deutsch-Jozsa algorithm. The proposed algorithm is based on the use of entanglement measure to explore 2n-2 additional classes compared to the standard Deutsch-Jozsa algorithm. In addition, the comparison between the proposed quantum algorithm and the classical one is investigated in details. The comparison shows that the proposed algorithm is faster when the number of Boolean variables exceed 14 variables.

Term evaluation metrics in imbalanced text categorization

AbstractThis paper proposes four novel term evaluation metrics to represent documents in the text categorization where class distribution is imbalanced. These metrics are achieved from the revision of the four common term evaluation metrics:chi-square,information gain,odds ratio, andrelevance frequency. While the common metrics require a balanced class distribution, our proposed metrics evaluate the document terms under an imbalanced distribution. They calculate the degree of relatedness of terms with respect to minor and major classes by considering their imbalanced distribution. Using these metrics in the document representation makes a better distinction between the documents of the minor and major classes and improves the performance of machine learning algorithms. The proposed metrics are assessed over three popular benchmarks (two subsets of Reuters-21578 and WebKB) by using four classification algorithms: support vector machines, naive Bayes, decision trees, and centroid-based classifiers. Our empirical results indicate that the proposed metrics outperform the common metrics in the imbalanced text categorization.

A Wide Scale Classification of Class Imbalance Problem and its Solutions: A Systematic Literature Review

In today’s world, most of the data (real world) is present in imbalanced form by nature. This is because of not having efficient algorithms to put this data (i.e., generated data by billion of internet- connected devices (IoTs)) in respective format. Imbalanced data poses a great challenge to (both) data mining and machine learning algorithms. The imbalanced dataset consists of a majority class and a minority class, where the majority class takes the lead over the minority class. Generally, several standard learning algorithms assume the balanced class distribution or equal misclassification costs. If prediction is performed by these learning algorithms on imbalanced data, the accuracy will be high for majority classes, i.e., resulting in poor performance. To overcome this problem (or improving accuracy of deision/prediction-making process), data mining and machine learning researchers have addressed the problem of imbalanced data using data-level, algorithmic level and ensemble or hybrid methods. This article presents a systematic literature review and analyze the results of more than 400 research papers published between 2002-2017 (till June 2017), resulting in a broader and elaborate investigation of the literature in this area of research. Note that extension of this article/work will contain till December 2018 research articles, which will be published in June 2019 (now these more papers/articles did not include due to no. of pages/space issues). The systematic analysis of the research literature has focus on the key role of Data Intrinsic Problems in classification, handling the imbalanced data and the techniques used to overcome the skewed distribution. Furthermore, this article reveals patterns, trends and gaps in the existing literature and discusses briefly the next generation research directions in this area.

Transfer synthetic over-sampling for class-imbalance learning with limited minority class data

The problem of limited minority class data is encountered in many class imbalanced applications, but has received little attention. Synthetic over-sampling, as popular class-imbalance learning methods, could introduce much noise when minority class has limited data since the synthetic samples are not i.i.d. samples of minority class. Most sophisticated synthetic sampling methods tackle this problem by denoising or generating samples more consistent with ground-truth data distribution. But their assumptions about true noise or ground-truth data distribution may not hold. To adapt synthetic sampling to the problem of limited minority class data, the proposed Traso framework treats synthetic minority class samples as an additional data source, and exploits transfer learning to transfer knowledge from them to minority class. As an implementation, TrasoBoost method firstly generates synthetic samples to balance class sizes. Then in each boosting iteration, the weights of synthetic samples and original data decrease and increase respectively when being misclassified, and remain unchanged otherwise. The misclassified synthetic samples are potential noise, and thus have smaller influence in the following iterations. Besides, the weights of minority class instances have greater change than those of majority class instances to be more influential. And only original data are used to estimate error rate to be immune from noise. Finally, since the synthetic samples are highly related to minority class, all of the weak learners are aggregated for prediction. Experimental results show TrasoBoost outperforms many popular class-imbalance learning methods.

Holomorphic Deformations of Balanced Calabi-Yau ∂∂ ¯-Manifolds

Given a compact complex n-fold X satisfying the ∂∂ ¯-lemma and supposed to have a trivial canonical bundle K X and to admit a balanced (=semi-Kähler) Hermitian metric ω, we introduce the concept of deformations of X that are co-polarised by the balanced class [ω n-1 ]∈H n-1,n-1 (X,ℂ)⊂H 2n-2 (X,ℂ) and show that the resulting theory of balanced co-polarised deformations is a natural extension of the classical theory of Kähler polarised deformations in the context of Calabi–Yau or holomorphic symplectic compact complex manifolds. The concept of Weil–Petersson metric still makes sense in this strictly more general, possibly non-Kähler context, while the Local Torelli Theorem still holds.

Profiling of local disease-sparing responses to bovine respiratory syncytial virus in intranasally vaccinated and challenged calves

Bovine and human respiratory syncytial viruses (BRSV, HRSV) are primary causes of pneumonia in calves and children respectively, with vaccination offering protection via antibody and cellular immune responses. However, with no vaccines currently licensed for human use, evaluation of local responses to BRSV vaccination may provide insights to aid the design of effective safe HRSV vaccines. Calves received intranasal single component BRSV vaccine or “3-Way” vaccine (BRSV, Bovine Herpes Virus-1 (BHV-1), Bovine Parainfluenza Virus Type-3 (BPIV-3)), and were BRSV-challenged 42 days post-vaccination. All vaccinates exhibited reduced pulmonary lesioning with elevated anti-BRSV serum IgG, and higher nasal anti-BRSV IgA in 3-Way vaccinates. Thirty-nine proteins associated with homeostatic and immune processes were altered in vaccinates, with enhanced 3-Way vaccinate group proteins associated with Th1/Th2 balance and immunoglobulin class switching. Proteins altered in the pharyngeal tonsil of animals euthanized early related to anti-inflammatory responses and lymphoid tissue remodeling. These findings indicate that multivalent vaccines distinctly modulate local immune responses, with clear correlation between the pharyngeal tonsil proteome profile and resulting immune protection and disease-sparing. This suggests that the efficacy of low-antigenic subunit vaccine components for problematic pathogens such as HRSV could be enhanced by use in combination with existing safe live vaccines. SignificanceThis study demonstrates that vaccine valency can alter post-challenge proteome responses within the pharyngeal tonsil, a sentinel site of primary immune responses, with the magnitude of response dependent on antigen formulation. Observed differential responses can be attributed to antigenic material and viral nucleic acid from multivalent formulations providing additional T-cell epitopes and PAMPS. These findings indicate that incorporation of subunits proteins within multivalent formulations containing live virus has the potential to induce/skew a favorable immune response, utilising the natural adjuvanting effects of safe proven live vaccines.

Deep Imbalanced Learning for Face Recognition and Attribute Prediction.

Data for face analysis often exhibit highly-skewed class distribution, i.e., most data belong to a few majority classes, while the minority classes only contain a scarce amount of instances. To mitigate this issue, contemporary deep learning methods typically follow classic strategies such as class re-sampling or cost-sensitive training. In this paper, we conduct extensive and systematic experiments to validate the effectiveness of these classic schemes for representation learning on class-imbalanced data. We further demonstrate that more discriminative deep representation can be learned by enforcing a deep network to maintain inter-cluster margins both within and between classes. This tight constraint effectively reduces the class imbalance inherent in the local data neighborhood, thus carving much more balanced class boundaries locally. We show that it is easy to deploy angular margins between the cluster distributions on a hypersphere manifold. Such learned Cluster-based Large Margin Local Embedding (CLMLE), when combined with a simple k-nearest cluster algorithm, shows significant improvements in accuracy over existing methods on both face recognition and face attribute prediction tasks that exhibit imbalanced class distribution.

Preoperative evaluation of tumour consistency in pituitary macroadenomas: a machine learning-based histogram analysis on conventional T2-weighted MRI.

To evaluate the potential value of machine learning (ML)-based histogram analysis (or first-order texture analysis) on T2-weighted magnetic resonance imaging (MRI) for predicting consistency of pituitary macroadenomas (PMA) and to compare it with that of signal intensity ratio (SIR) evaluation. Fifty-five patients with 13 hard and 42 soft PMAs were included in this retrospective study. Histogram features were extracted from coronal T2-weighted original, filtered and transformed MRI images by manual segmentation. To achieve balanced classes (38 hard vs 42 soft), multiple samples were obtained from different slices of the PMAs with hard consistency. Dimension reduction was done with reproducibility analysis, collinearity analysis and feature selection. ML classifier was artificial neural network (ANN). Reference standard for the classifications was based on surgical and histopathological findings. Predictive performance of histogram analysis was compared with that of SIR evaluation. The main metric for comparisons was the area under the receiver operating characteristic curve (AUC). Only 137 of 162 features had excellent reproducibility. Collinearity analysis yielded 20 features. Feature selection algorithm provided six texture features. For histogram analysis, the ANN correctly classified 72.5% of the PMAs regarding consistency with an AUC value of 0.710. For SIR evaluation, accuracy and AUC values were 74.5% and 0.551, respectively. Considering AUC values, ML-based histogram analysis performed better than SIR evaluation (z = 2.312, p = 0.021). ML-based T2-weighted MRI histogram analysis might be a useful technique in predicting the consistency of PMAs, with a better predictive performance than that of SIR evaluation.

Supervised Machine-Learning Predictive Analytics for National Quality of Life Scoring

For many years there has been a focus on individual welfare and societal advancement. In addition to the economic system, diverse experiences and the habitats of people are crucial factors that contribute to the well-being and progress of the nation. The predictor of quality of life called the Better Life Index (BLI) visualizes and compares key elements—environment, jobs, health, civic engagement, governance, education, access to services, housing, community, and income—that contribute to well-being in different countries. This paper presents a supervised machine-learning analytical model that predicts the life satisfaction score of any specific country based on these given parameters. This work is a stacked generalization based on a novel approach that combines different machine-learning approaches to generate a meta-machine-learning model that further aids in maximizing prediction accuracy. The work utilized an Organization for Economic Cooperation and Development (OECD) regional statistics dataset with four years of data, from 2014 to 2017. The novel model achieved a high root mean squared error (RMSE) value of 0.3 with 10-fold cross-validation on the balanced class data. Compared to base models, the ensemble model based on the stacked generalization framework was a significantly better predictor of the life satisfaction of a nation. It is clear from the results that the ensemble model presents more precise and consistent predictions in comparison to the base learners.

Real-value negative selection over-sampling for imbalanced data set learning

The learning problem from imbalanced data set poses a major challenge in data mining community. Conventional machine learning algorithms show poor performance in dealing with the classification problems of imbalanced data set since they are originally designed to work with balanced class distribution. In this paper, we propose a new over-sampling technique, which uses the real-value negative selection (RNS) procedure to generate artificial minority data with no requirement of actual minority data available. The generated minority data with rare actual minority data if available are combined with the majority data as input to a bi-class classification approach for learning. In the experiments, we demonstrate the effectiveness of RNS in avoiding the problems often encountered by the existing over-sampling methods such as the generation of noisy instances and almost duplicated instances in the same clusters. Moreover, the extensive experimental results on the different imbalanced datasets from UCI repository and real-world imbalanced datasets show that when dealing with the classification of imbalanced datasets, the proposed hybrid approach can achieve better performance in terms of both G-Mean and F-Measure evaluation metrics as compared to the other existing imbalanced dataset classification techniques.

Gimme shelter? Heterogeneous preferences for tangible and intangible resources when choosing an incubator

The rapidly growing and diversifying incubator population has led to increasing efforts to understand why entrepreneurs prefer one incubator over another. Scientific studies suggest that entrepreneurs should prefer incubators that provide startups with intangible resources, such as business knowledge or networks to enhance performance. Yet, studies show many entrepreneurs prefer incubators that provide tangible resources, such as funding and office space. The heterogeneity in preferences for resources from incubators is poorly understood. We do not know whether there are patterns in this heterogeneity nor what factors explain this heterogeneity. Thereby, we do not know the extent to which a one-size-fits-all model of incubation is sufficient to attract and support startups or whether incubators need to tailor themselves to the perceived resource needs of different groups of startups. Hence, this paper aims to identify and explain the heterogeneity in preferences for resources offered by an incubator to startups. We conducted a discrete choice experiment to determine how the attributes of an incubator influence incubator choice by different latent classes of entrepreneurs. The data comes from 935 entrepreneurs in North America and Western Europe. Our results reveal three latent classes of entrepreneurs: “ambitious, balanced spinoffs,” who consider all the incubator’s attributes when making a decision; “innovation-driven funding seekers,” who base their choice on funding availability; and “self-made individualists,” who disfavor networking, training, and coaching. The ambitious, balanced spinoffs class based their choice on the attributes highlighted in the literature, while the innovation-driven funding seekers and self-made individualists fit more with empirically observed preferences for tangible resources. The classes show that systematic heterogeneity exists in the preferences for resources provided by an incubator. We advise incubator managers how to better tailor their support to the preferences and needs of classes of startups.

Effects of a falls exercise intervention on strength, power, functional ability and bone in older frequent fallers: FaME (Falls Management Exercise) RCT secondary analysis.

Objectives:Falls Management Exercise (FaME) has been shown to reduce falls in frequent fallers and in lower risk sedentary older people. The effects of FaME on the strength, power, physical function and bone health of frequently falling older women are yet to be established.Methods:This paper reports secondary analysis of data from the original randomised controlled trial of FaME in 100 community dwelling women aged ≥65 years with a history of ≥3 falls in the previous year. Intervention was group delivered, weekly one hour tailored dynamic balance and strength exercise classes and home exercise for nine months.Outcome measures included:strength (handgrip, quadriceps, hamstrings, hip abductors, ankles), lower limb explosive power and functional tests (timed up and go, functional reach, timed floor rise and balance), analysed using Linear Mixed Model analysis. Bone Mineral Density (BMD) at hip and spine was measured in a smaller sub-group and analysed using t-tests.Results:Significant time*group interactions in all measures of strength, except isometric ankle dorsiflexion, concentric hamstring and eccentric quadriceps strength. These improvements in strength equated to average improvements of 7-45%. There were also significant improvements in explosive power (W/kg) (18%, p=0.000), timed up and go (16%, p=0.000), functional reach (17%, p=0.000), floor rise (10%, p=0.002) and eyes closed static balance (56%, p=0.000). There was a significant loss of hip BMD in the control group (neck of femur p<0.05; ward’s triangle p<0.02).Conclusion:The FaME intervention improves lower limb strength, power and clinically relevant functional outcomes in frequently falling older women.

Boosting the performance of over-sampling algorithms through under-sampling the minority class

Over-sampling algorithms are the most adopted approach to balance class distribution in imbalanced data problems, through random replication or synthesis of new examples in the minority class. Current over-sampling algorithms, however, usually use all available examples in the minority class to synthesise new instances, which may include noisy or outlier data. This work proposes k-INOS, a new algorithm to prevent over-sampling algorithms from being contaminated by noisy examples in the minority class. k-INOS is based on the concept of neighbourhood of influence and works as a wrapper around any over-sampling algorithm. A comprehensive experimentation was conducted to test k-INOS in 50 benchmark data sets, 8 over-sampling methods and 5 classifiers, with performance measured according to 7 metrics and Wilcoxon signed-ranks test. Results showed, particularly for weak classifiers (but not only), k-INOS significantly improved the performance of over-sampling algorithms in most performance metrics. Further investigations also allowed to identify conditions where k-INOS is likely to increase performance, according to features and rates measured from the data sets. The extensive experimentation framework evidenced k-INOS as an efficient algorithm to be applied prior to over-sampling methods.

Radiogenomics in Clear Cell Renal Cell Carcinoma: Machine Learning-Based High-Dimensional Quantitative CT Texture Analysis in Predicting PBRM1 Mutation Status.

The purpose of this study is to evaluate the potential value of machine learning (ML)-based high-dimensional quantitative CT texture analysis in predicting the mutation status of the gene encoding the protein polybromo-1 (PBRM1) in patients with clear cell renal cell carcinoma (RCC). In this retrospective study, 45 patients with clear cell RCC (29 without the PBRM1 mutation and 16 with the PBRM1 mutation) were identified in The Cancer Genome Atlas-Kidney Renal Clear Cell Carcinoma database. To create stable ML models and balanced classes, the data were augmented to a total of 161 labeled segmentations (87 without the PBRM1 mutation and 74 with the PBRM1 mutation) by obtaining three to five different samples per patient. Texture features were extracted from corticomedullary phase contrast-enhanced CT images with the use of an open-source software package for the extraction of radiomic data from medical images. Reproducibility analysis (intraclass correlation) was performed by two radiologists. Attribute selection and model optimization were done using a wrapper-based classifier-specific algorithm with nested cross-validation. ML classifiers were an artificial neural network (ANN) algorithm and a random forest (RF) algorithm. The models were validated using 10-fold cross-validation. The reference standard was the PBRM1 mutation status. The main performance metric was the AUC value. Of 828 extracted texture features, 759 had excellent reproducibility. Using 10 selected features, the ANN algorithm correctly classified 88.2% (142 of 161) of the clear cell RCCs in terms of PBRM1 mutation status (AUC value, 0.925). Using five selected features, the RF algorithm correctly classified 95.0% (153 of 161) of the clear cell RCCs (AUC value, 0.987). Overall, the RF algorithm performed better than the ANN algorithm (z score = -2.677; p = 0.007). ML-based high-dimensional quantitative CT texture analysis might be a feasible and potential method for predicting PBRM1 mutation status in patients with clear cell RCC.

Integration of synthetic minority oversampling technique for imbalanced class

In the data mining, a class imbalance is a problematic issue to look for the solutions. It probably because machine learning is constructed by using algorithms with assuming the number of instances in each balanced class, so when using a class imbalance, it is possible that the prediction results are not appropriate. They are solutions offered to solve class imbalance issues, including oversampling, undersampling, and synthetic minority oversampling technique (SMOTE). Both oversampling and undersampling have its disadvantages, so SMOTE is an alternative to overcome it. By integrating SMOTE in the data mining classification method such as Naive Bayes, Support Vector Machine (SVM), and Random Forest (RF) is expected to improve the performance of accuracy. In this research, it was found that the data of SMOTE gave better accuracy than the original data. In addition to the three classification methods used, RF gives the highest average AUC, F-measure, and G-means score.

Efficiency of ODI and APDI of Kim's cephalometric analysis in a Latin American population with skeletal open bite.

ABSTRACTObjective: The objective of this research was to demonstrate the efficiency of the overbite depth indicator (ODI) and the anteroposterior dysplasia indicator (APDI) from Kim’s cephalometric analysis, regarding the determination of the vertical and sagittal patterns of Latin American individuals. Methods: Two hundred lateral cephalometric radiographs were selected and divided into four study groups, with 50 radiographs each, for carrying out a cross-sectional study. The control group included radiographs of balanced individuals, and the other three groups had lateral cephalometric radiographs of subjects with Class I, II and III malocclusions and with skeletal open bite. After the pilot test was performed to calibrate the investigator, the ODI and APDI were measured. Descriptive statistics were performed and the one-way ANOVA with post-hoc Tukey HSD, or Kruskal-Wallis and Mann-Whitney U-test were used. Also a multiple linear regression was employed.Results: Statistically significant differences were found for the ODI of all groups (p< 0.001), except between Class I group (65.87 ± 4.26) and Class II open bite group (67.19 ± 3.58), both with similar values to each other. For APDI, statistically significant differences were also found for all groups (p< 0.001). However, no statistically significant differences were found between the balanced group (83.18 ± 1.71) and Class I group with skeletal open bite (81.78 ± 2.69). Conclusions: ODI and APDI are reliable indicators to evaluate the sagittal and vertical patterns of an individual, demonstrating their efficiency when a Latin American population was evaluated.

VCOS: A Novel Synergistic Oversampling Algorithm in Binary Imbalance Classification

Learning from class-imbalanced data is a challenging problem as standard classification algorithms are designed to handle balanced class distributions. Scholars solve this problem by modifying classifiers or and generating artificial data by oversampling. The former usually design corresponding classifier to adapt them to the imbalanced data, while the latter exploits the sampling algorithm, which are the data preprocessing steps independent of the classifier. In this paper, we propose a novel synergistic oversampling algorithm to combine the oversampling and classification into one without training the classifier repeatedly, which can generate new pertinent samples according to the classification performance of the classifier without repeat training or deep understanding of the classifier, so the generated samples can guarantee the performance improvement of the classifier. Moreover, The proposed framework enclosures the oversampling method without traditional parameters in oversampling methods. Experimental results on several real-life imbalanced datasets demonstrate the effectiveness and efficiency of the proposed algorithm in binary classification problems.

Variance Ranking Attributes Selection Techniques for Binary Classification Problem in Imbalance Data

Data are being generated and used to support all aspects of healthcare provision, from policy formation to the delivery of primary care services. Particularly, with the change of emphasis from curative to preventive medicine, the importance of data-based research such as data mining and machine learning has emphasized the issues of class distributions in datasets. In typical predictive modeling, the inability to effectively address a class imbalance in a real-life dataset is an important shortcoming of the existing machine learning algorithms. Most algorithms assume a balanced class in their design, resulting in poor performance in predicting the minority target class. Ironically, the minority target class is usually the focus in predicting processes. The misclassification of the minority target class has resulted in serious consequences in detecting chronic diseases and detecting fraud and intrusion where positive cases are erroneously predicted as not positive. This paper presents a new attribute selection technique called variance ranking for handling imbalance class problems in a dataset. The results obtained were compared to two well-known attribute selection techniques: the Pearson correlation and information gain technique. This paper uses a novel similarity measurement technique ranked order similarity-ROS to evaluate the variance ranking attribute selection compared to the Pearson correlations and information gain. Further validation was carried out using three binary classifications: logistic regression, support vector machine, and decision tree. The proposed variance ranking and ranked order similarity techniques showed better results than the benchmarks. The ROS technique provided an excellent means of grading and measuring the similarities where other similarity measurement techniques were inadequate or not applicable.

Factors Associated With Older Adults' Enrollment in Balance Classes to Prevent Falls: Case-Control Study.

Balance training decreases fall risk among older adults, but few participate in such training. We examined the association of exposure to social marketing to promote balance classes, personal characteristics and other factors, with older adults' balance class participation. Adults aged ⩾60 years were eligible for this case-control study if they attended any church enrolled in a trial testing the effect of social marketing on balance class participation. Cases attended balance classes during the study period; controls were randomly sampled congregants who did not join a class. Cases were more likely to attend churches that received the social marketing program, and were older, more often female, and more frequently experienced "near falls" than controls. Participation was also associated with increasing age of the church's leader and rural church attendance. Programs to promote balance classes may need to be tailored to target some risk groups, including men and urban and suburban congregants.