FF-SMOTE: A Metaheuristic Approach to Combat Class Imbalance in Binary Classification

Nature inspired intelligent computation-based algorithms have been grown remarkably over the previous years. These algorithms are applied for optimizing the values to a number of problem areas ranging from scientific research to industry or commerce. Class imbalance is a challenging problem of classification to identify smaller class when dealing with skewed distributions. This paper proposed a firefly-based oversampling technique to combat class imbalance in binary classification. The proposed technique is applied on 10 UCI data sets with the imbalance ratio ranging from high to low and is compared with the other state-of-the art oversampling techniques. The performance of the proposed method is assessed through performance metrics area under the curve and geometric mean. The techniques are also analyzed statistically using Friedman and Wilcoxon matched signed rank Test. Through experimental and statistical analysis, it is reported that the proposed technique outperformed other oversampling techniques.