Undersampled $$K$$ K -means approach for handling imbalanced distributed data

Influence of resampling techniques on Bayesian network performance in predicting increased algal activity

Early warning of increased algal activity is important to mitigate potential impacts on aquatic life and human health. While many methods have been developed to predict increased algal activity, an ongoing issue is that severe algal blooms often occur with low frequency in water bodies. This results in imbalanced data sets available for model specification, leading to poor predictions of the frequency of increased algal activity. One approach to address this is to resample data sets of increased algal activity to increase the prevalence of higher than normal algal activity in calibration data and ultimately improve model predictions. This study aims to investigate the use of resampling techniques to address the imbalanced dataset and determine if such methods can improve the prediction of increased algal activity. Three techniques were investigated, Kmeans under-sampling (US_Kmeans), synthetic minority over-sampling technique (SMOTE), and 'SMOTE and cluster-based under-sampling technique' (SCUT). The resampling methods were applied to a Bayesian network (BN) model of Lake Burragorang in New South Wales, Australia. The model was developed to predict chlorophyll-a (chl-a) using a range of water quality parameters as predictors. The original data and each of the balanced datasets were used for BN structures and parameter learning. The results showed that the best graphical structure was obtained by adding synthetic data from SMOTE with the highest true positive rate (TPR) and area under the curve (AUC). When compared using a fixed graphical structure for the BN, all resampling techniques increased the ability of the BN to detect events with higher probability of increased algal activity. The resampling model results can also be used to better understand the most important influences on high chl-a concentrations and suggest future data collection and model development priorities.

A novel multi-stage ensemble model for credit scoring based on synthetic sampling and feature transformation

Credit scoring has become increasingly important for financial institutions. With the advancement of artificial intelligence, machine learning methods, especially ensemble learning methods, have become increasingly popular for credit scoring. However, the problems of imbalanced data distribution and underutilized feature information have not been well addressed sufficiently. To make the credit scoring model more adaptable to imbalanced datasets, the original model-based synthetic sampling method is extended herein to balance the datasets by generating appropriate minority samples to alleviate class overlap. To enable the credit scoring model to extract inherent correlations from features, a new bagging-based feature transformation method is proposed, which transforms features using a tree-based algorithm and selects features using the chi-square statistic. Furthermore, a two-layer ensemble method that combines the advantages of dynamic ensemble selection and stacking is proposed to improve the classification performance of the proposed multi-stage ensemble model. Finally, four standardized datasets are used to evaluate the performance of the proposed ensemble model using six evaluation metrics. The experimental results confirm that the proposed ensemble model is effective in improving classification performance and is superior to other benchmark models.

Spatio-temporal modeling of PM2.5 risk mapping using three machine learning algorithms

Urban air pollution is one of the most critical issues that affect the environment, community health, economy, and management of urban areas. From a public health perspective, PM_2.5 is one of the primary air pollutants, especially in Tehran's metropolis. Owing to the different patterns of PM_2.5 in different seasons, Spatio-temporal modeling and identification of high-risk areas to reduce its effects seems necessary. The purpose of this study was Spatio-temporal modeling and preparation of PM_2.5 risk mapping using three machine learning algorithms (random forest (RF), AdaBoost, and stochastic gradient descent (SGD)) in the metropolis of Tehran, Iran. Therefore, in the first step, to prepare the dependent variable data, the PM_2.5 average was used for the four seasons of spring, summer, autumn, and winter. Then, using remote sensing (RS) and a geographic information system (GIS), independent data such as temperature, maximum temperature, minimum temperature, wind speed, rainfall, humidity, normalized difference vegetation index (NDVI), population density, street density, and distance to industrial centers were prepared as a seasonal average. To Spatio-temporal modeling using machine learning algorithms, 70% of the data were used for training and 30% for validation. The frequency ratio (FR) model was used as input to machine learning algorithms to calculate the spatial relationship between PM_2.5 and the effective parameters. Finally, Spatio-temporal modeling and PM_2.5 risk mapping were performed using three machine learning algorithms. The receiver operating characteristic (ROC) area under the curve (AUC) results showed that the RF algorithm had the greatest modeling accuracy, with values of 0.926, 0.94, 0.949, and 0.949 for spring, summer, autumn, and winter, respectively. According to the RF model, the most important variable in spring and autumn was NDVI. Temperature and distance to industrial centers were the most important variables in the summer and winter, respectively. The results showed that autumn, winter, summer, and spring had the highest risk of PM_2.5, respectively.

GHOST: Adjusting the Decision Threshold to Handle Imbalanced Data in Machine Learning

Machine learning classifiers trained on class imbalanced data are prone to overpredict the majority class. This leads to a larger misclassification rate for the minority class, which in many real-world applications is the class of interest. For binary data, the classification threshold is set by default to 0.5 which, however, is often not ideal for imbalanced data. Adjusting the decision threshold is a good strategy to deal with the class imbalance problem. In this work, we present two different automated procedures for the selection of the optimal decision threshold for imbalanced classification. A major advantage of our procedures is that they do not require retraining of the machine learning models or resampling of the training data. The first approach is specific for random forest (RF), while the second approach, named GHOST, can be potentially applied to any machine learning classifier. We tested these procedures on 138 public drug discovery data sets containing structure-activity data for a variety of pharmaceutical targets. We show that both thresholding methods improve significantly the performance of RF. We tested the use of GHOST with four different classifiers in combination with two molecular descriptors, and we found that most classifiers benefit from threshold optimization. GHOST also outperformed other strategies, including random undersampling and conformal prediction. Finally, we show that our thresholding procedures can be effectively applied to real-world drug discovery projects, where the imbalance and characteristics of the data vary greatly between the training and test sets.

An Impartial Semi-Supervised Learning Strategy for Imbalanced Classification on VHR Images

Imbalanced learning is a common problem in remote sensing imagery-based land-use and land-cover classifications. Imbalanced learning can lead to a reduction in classification accuracy and even the omission of the minority class. In this paper, an impartial semi-supervised learning strategy based on extreme gradient boosting (ISS-XGB) is proposed to classify very high resolution (VHR) images with imbalanced data. ISS-XGB solves multi-class classification by using several semi-supervised classifiers. It first employs multi-group unlabeled data to eliminate the imbalance of training samples and then utilizes gradient boosting-based regression to simulate the target classes with positive and unlabeled samples. In this study, experiments were conducted on eight study areas with different imbalanced situations. The results showed that ISS-XGB provided a comparable but more stable performance than most commonly used classification approaches (i.e., random forest (RF), XGB, multilayer perceptron (MLP), and support vector machine (SVM)), positive and unlabeled learning (PU-Learning) methods (PU-BP and PU-SVM), and typical synthetic sample-based imbalanced learning methods. Especially under extremely imbalanced situations, ISS-XGB can provide high accuracy for the minority class without losing overall performance (the average overall accuracy achieves 85.92%). The proposed strategy has great potential in solving the imbalanced classification problems in remote sensing.

A cluster-based hybrid sampling approach for imbalanced data classification

When processing instrumental data by using classification approaches, the imbalanced dataset problem is usually challenging. As the minority class instances could be overwhelmed by the majority class instances, training a typical classifier with such a dataset directly might get poor results in classifying the minority class. We propose a cluster-based hybrid sampling approach CUSS (Cluster-based Under-sampling and SMOTE) for imbalanced dataset classification, which belongs to the type of data-level methods and is different from previously proposed hybrid methods. A new cluster-based under-sampling method is designed for CUSS, and a new strategy to set the expected instance number according to data distribution in the original training dataset is also proposed in this paper. The proposed method is compared with five other popular resampling methods on 15 datasets with different instance numbers and different imbalance ratios. The experimental results show that the CUSS method has good performance and outperforms other state-of-the-art methods.

Regression on imperfect class labels derived by unsupervised clustering

Outcome regressed on class labels identified by unsupervised clustering is custom in many applications. However, it is common to ignore the misclassification of class labels caused by the learning algorithm, which potentially leads to serious bias of the estimated effect parameters. Due to their generality we suggest to address the problem by use of regression calibration or the misclassification simulation and extrapolation method. Performance is illustrated by simulated data from Gaussian mixture models, documenting a reduced bias and improved coverage of confidence intervals when adjusting for misclassification with either method. Finally, we apply our method to data from a previous study, which regressed overall survival on class labels derived from unsupervised clustering of gene expression data from bone marrow samples of multiple myeloma patients.