Existing classification analysis techniques (e.g., decision tree induction,) generally exhibit satisfactory classification effectiveness when dealing with data with non-skewed class distribution. However, real-world applications (e.g., churn prediction and fraud detection) often involve highly skewed data in decision outcomes. Such a highly skewed class distribution problem, if not properly addressed, would imperil the resulting learning effectiveness. In this study, we empirically evaluate three different approaches, namely the under-sampling, the over-sampling and the multi-classifier committee approaches, for addressing classification with highly skewed class distribution. Due to its popularity, C4.5 is selected as the underlying classification analysis technique. Based on 10 highly skewed class distribution datasets, our empirical evaluations suggest that the multi-classifier committee generally outperformed the under-sampling and the over-sampling approaches, using the recall rate, precision rate and F1-measure as the evaluation criteria. Furthermore, for applications aiming at a high recall rate, use of the over-sampling approach will be suggested. On the other hand, if the precision rate is the primary concern, adoption of the classification model induced directly from original datasets would be recommended.

Abstract Existing classification analysis techniques (e.g., decision tree induction, backpropagation neural network, k-nearest neighbor classification, etc.) generally exhibit satisfactory classification effectiveness when dealing with data with non-skewed class distribution. However, real-world applications (e.g., churn prediction and fraud detection) often involve highly skewed data in decision outcomes (e.g., 2% churners and 98% non-churners). Such a highly skewed class distribution problem, if not properly addressed, would imperil the resulting learning effectiveness and might result in a ¡§null¡¨ prediction system that simply predicts all instances as having the majority decision class as the training instances (e.g., predicting all customers as non-churners). In this study, we extended the multi-classifier class-combiner approach and proposed a clustering-based multi-classifier class-combiner technique to address the highly skewed class distribution problem in classification analysis. In addition, we proposed four distance-based methods for selecting a subset of instances having the majority decision class for lowering the degree of skewness in a data set. Using two real-world datasets (including mortality prediction for burn patients and customer loyalty prediction), empirical results suggested that the proposed clustering-based multi-classifier class-combiner technique generally outperformed the traditional multi-classifier class-combiner approach and the four distance-based methods. Keywords: Data Mining, Classification Analysis, Skewed Class Distribution Problem, Decision Tree Induction, Multi-classifier Class-combiner Approach, Clustering-based Multi-classifier Class-combiner Approach