A dynamic self-adoptive genetic algorithm for personal credit risk assessment

Simple genetic algorithm has many defects, such as premature and slow speed of convergence. This paper researches the frame and performance of four combination algorithms based on dynamic self-adaptive genetic algorithm (DSGA-SVM, DSGA-Logistic, DSGA-C4.5, DSGA-BPNN). In order to classify the customers into two groups representing low and high credit risk, the proposed algorithms are tested using three countries´ personal credit data download from the website of UCI machine learning. Through the comparison of the algorithms proposed above we can verify the performance of DSGA-based algorithms and check out the most suitable algorithms to combine with DSGA.