A hierarchical genetic fuzzy system based on genetic programming for addressing classification with highly imbalanced and borderline data-sets

A Boundary-Information-Based Oversampling Approach to Improve Learning Performance for Imbalanced Datasets

Oversampling is the most popular data preprocessing technique. It makes traditional classifiers available for learning from imbalanced data. Through an overall review of oversampling techniques (oversamplers), we find that some of them can be regarded as danger-information-based oversamplers (DIBOs) that create samples near danger areas to make it possible for these positive examples to be correctly classified, and others are safe-information-based oversamplers (SIBOs) that create samples near safe areas to increase the correct rate of predicted positive values. However, DIBOs cause misclassification of too many negative examples in the overlapped areas, and SIBOs cause incorrect classification of too many borderline positive examples. Based on their advantages and disadvantages, a boundary-information-based oversampler (BIBO) is proposed. First, a concept of boundary information that considers safe information and dangerous information at the same time is proposed that makes created samples near decision boundaries. The experimental results show that DIBOs and BIBO perform better than SIBOs on the basic metrics of recall and negative class precision; SIBOs and BIBO perform better than DIBOs on the basic metrics for specificity and positive class precision, and BIBO is better than both of DIBOs and SIBOs in terms of integrated metrics.

A Review of Fuzzy and Pattern-Based Approaches for Class Imbalance Problems

The usage of imbalanced databases is a recurrent problem in real-world data such as medical diagnostic, fraud detection, and pattern recognition. Nevertheless, in class imbalance problems, the classifiers are commonly biased by the class with more objects (majority class) and ignore the class with fewer objects (minority class). There are different ways to solve the class imbalance problem, and there has been a trend towards the usage of patterns and fuzzy approaches due to the favorable results. In this paper, we provide an in-depth review of popular methods for imbalanced databases related to patterns and fuzzy approaches. The reviewed papers include classifiers, data preprocessing, and evaluation metrics. We identify different application domains and describe how the methods are used. Finally, we suggest further research directions according to the analysis of the reviewed papers and the trend of the state of the art.

Multiset Feature Learning for Highly Imbalanced Data Classification

With the expansion of data, increasing imbalanced data has emerged. When the imbalance ratio (IR) of data is high, most existing imbalanced learning methods decline seriously in classification performance. In this paper, we systematically investigate the highly imbalanced data classification problem, and propose an uncorrelated cost-sensitive multiset learning (UCML) approach for it. Specifically, UCML first constructs multiple balanced subsets through random partition, and then employs the multiset feature learning (MFL) to learn discriminant features from the constructed multiset. To enhance the usability of each subset and deal with the non-linearity issue existed in each subset, we further propose a deep metric based UCML (DM-UCML) approach. DM-UCML introduces the generative adversarial network technique into the multiset constructing process, such that each subset can own similar distribution with the original dataset. To cope with the non-linearity issue, DM-UCML integrates deep metric learning with MFL, such that more favorable performance can be achieved. In addition, DM-UCML designs a new discriminant term to enhance the discriminability of learned metrics. Experiments on eight traditional highly class-imbalanced datasets and two large-scale datasets indicate that: the proposed approaches outperform state-of-the-art highly imbalanced learning methods and are more robust to high IR.

Computational model for vitamin D deficiency using hair mineral analysis

Vitamin D deficiency is prevalent in the Arabian Gulf region, especially among women. Recent studies show that the vitamin D deficiency is associated with a mineral status of a patient. Therefore, it is important to assess the mineral status of the patient to reveal the hidden mineral imbalance associated with vitamin D deficiency. A well-known test such as the red blood cells is fairly expensive, invasive, and less informative. On the other hand, a hair mineral analysis can be considered an accurate, excellent, highly informative tool to measure mineral imbalance associated with vitamin D deficiency. In this study, 118 apparently healthy Kuwaiti women were assessed for their mineral levels and vitamin D status by a hair mineral analysis (HMA). This information was used to build a computerized model that would predict vitamin D deficiency based on its association with the levels and ratios of minerals. The first phase of the proposed model introduces a novel hybrid optimization algorithm, which can be considered as an improvement of Bat Algorithm (BA) to select the most discriminative features. The improvement includes using the mutation process of Genetic Algorithm (GA) to update the positions of bats with the aim of speeding up convergence; thus, making the algorithm more feasible for wider ranges of real-world applications. Due to the imbalanced class distribution in our dataset, in the second phase, different sampling methods such as Random Under-Sampling, Random Over-Sampling, and Synthetic Minority Oversampling Technique are used to solve the problem of imbalanced datasets. In the third phase, an AdaBoost ensemble classifier is used to predicting the vitamin D deficiency. The results showed that the proposed model achieved good results to detect the deficiency in vitamin D.

BSFS: Design and Development of Exponential Brain Storm Fuzzy System for Data Classification

The inductive learning of fuzzy rule classifier suffers in the rule generation and rule optimization when the search space or variables becomes high. This creates the new idea of making the fuzzy system with precise rules leading to less scalability and improved accuracy. Accordingly, different approaches have been presented in the literature for optimal finding of fuzzy rules using optimization algorithms. Here, we make use of the brain storm optimization algorithm for rule optimization. In this paper, a new fuzzy system called, exponential brain storm fuzzy system is developed by modifying the traditional fuzzy system in rule definition process. In rule derivation, we have presented an algorithm called, EBSO by modifying the BSO algorithm with exponential model. Also, the membership function is designed using simple uniform distribution-based approach. Finally, data classification is performed with a new BSFS system using three medical databases such as, PID, Cleveland and DRD. The experimentation proved that the proposed BSFS clearly outperformed in all the three datasets by reaching the maximum accuracy.

A Predictive Model for Toxicity Effects Assessment of Biotransformed Hepatic Drugs Using Iterative Sampling Method

Measuring toxicity is one of the main steps in drug development. Hence, there is a high demand for computational models to predict the toxicity effects of the potential drugs. In this study, we used a dataset, which consists of four toxicity effects:mutagenic, tumorigenic, irritant and reproductive effects. The proposed model consists of three phases. In the first phase, rough set-based methods are used to select the most discriminative features for reducing the classification time and improving the classification performance. Due to the imbalanced class distribution, in the second phase, different sampling methods such as Random Under-Sampling, Random Over-Sampling and Synthetic Minority Oversampling Technique are used to solve the problem of imbalanced datasets. ITerative Sampling (ITS) method is proposed to avoid the limitations of those methods. ITS method has two steps. The first step (sampling step) iteratively modifies the prior distribution of the minority and majority classes. In the second step, a data cleaning method is used to remove the overlapping that is produced from the first step. In the third phase, Bagging classifier is used to classify an unknown drug into toxic or non-toxic. The experimental results proved that the proposed model performed well in classifying the unknown samples according to all toxic effects in the imbalanced datasets.

Modeling urban air pollution with optimized hierarchical fuzzy inference system

Environmental exposure assessments (EEA) and epidemiological studies require urban air pollution models with appropriate spatial and temporal resolutions. Uncertain available data and inflexible models can limit air pollution modeling techniques, particularly in under developing countries. This paper develops a hierarchical fuzzy inference system (HFIS) to model air pollution under different land use, transportation, and meteorological conditions. To improve performance, the system treats the issue as a large-scale and high-dimensional problem and develops the proposed model using a three-step approach. In the first step, a geospatial information system (GIS) and probabilistic methods are used to preprocess the data. In the second step, a hierarchical structure is generated based on the problem. In the third step, the accuracy and complexity of the model are simultaneously optimized with a multiple objective particle swarm optimization (MOPSO) algorithm. We examine the capabilities of the proposed model for predicting daily and annual mean PM2.5 and NO2 and compare the accuracy of the results with representative models from existing literature. The benefits provided by the model features, including probabilistic preprocessing, multi-objective optimization, and hierarchical structure, are precisely evaluated by comparing five different consecutive models in terms of accuracy and complexity criteria. Fivefold cross validation is used to assess the performance of the generated models. The respective average RMSEs and coefficients of determination (R (2)) for the test datasets using proposed model are as follows: daily PM2.5 = (8.13, 0.78), annual mean PM2.5 = (4.96, 0.80), daily NO2 = (5.63, 0.79), and annual mean NO2 = (2.89, 0.83). The obtained results demonstrate that the developed hierarchical fuzzy inference system can be utilized for modeling air pollution in EEA and epidemiological studies.