Financial crisis prediction (FCP) is highly essential for financial firms, which intends to eliminate the future losses. In the last decade, numerous FCP models are developed by utilizing the past financial data of a company. As the financial data is high-dimensional and holds many unnecessary features which can influence the overall classification accuracy. To avoid this problem, feature selection methods are used earlier to the data classification process. This paper develops an FCP model which employs genetic algorithm (GA) for feature selection in prior to classification process. The inclusion of GA-FS method eliminates the irrelevant features and helps to enhance the accuracy of the FCP model. The proposed FCP model is tested on two benchmark dataset namely qualitative bankruptcy dataset in addition to Australian Credit dataset. For comparison purposes, GA, CFS and CSF based feature selection methodologies are used. To test the classifier results, GA-FS method is resemblance with renowned logistic regression (LR) as well as radial basis function (RBF) classifiers. The results reveal that the proposed FCP model shows superior classification performance when comparingby precision, recall and F-score.
Volume 11 | 03-Special Issue
Pages: 1412-1420