CFR: collaborative feature ranking for improving the performance of credit scoring data classification

Credit scoring is a prominent research problem as its predictive performance is accountable for the viability of financial industry. Credit scoring datasets are high-dimensional which are related to customers’ credentials such as annual income, job status, residential status, etc. In high-dimensional data, many of the features may be irrelevant or redundant which leads to problems such as over-fitting and high computational overhead. So, effective feature selection approaches may overcome both the problems related to high-dimensional data. Generally, a features set selected by a feature selection approach is appropriate with a classifier not with all classifiers, and improves the classification performances. In this article, we have proposed a collaborative feature ranking approach with consideration of various measures which can improve the classification performance of most of the classifiers. Further, it is applied in five credit scoring datasets, results of the proposed approach are compared with various existing feature ranking approaches in terms of similarities between features selected by each pair of approaches and classification performances with respect to these sets of features.