Designing of a Risk Assessment Model for Issuing Credit Card Using Parallel Social Spider Algorithm

The financial creditability of the customer needs to be verified by the lender/bank before issuing a credit card. This involves assessment of factors like the economic, social or social-economic background of the person. Thus, the features incorporated into the analysis are mixed data type ex. Income (numerical) and Property Owned (Categorical). In this manuscript, a credit card lending model is designed using a recently proposed parallel social spider algorithm by Shukla and Nanda in 2016. Suitable modifications have been introduced in the coding scheme and mating procedure to efficiently solve the credit assessment problem. Experiments are carried out on various standard credit card data available like German, Australian and Japanese credit card datasets. The superior performance of proposed algorithm is reported as compared to that achieved by K-means, parallel real genetic algorithm and parallel particle swarm optimization (PPSO). The Silhouette Index obtained by various algorithms specifically for Germen dataset are 0.56% by K-means, 0.86% by parallel Real Coded Genetic algorithm, 0.71% by PPSO and 0.84% by proposed method.