A Genetic Algorithm Based Classification for Credit Risk Assessment Problems

Credit risk assessment refers to the methods of which banks and financial institutions adopt to measure the potential of borrowers (customers) to repay loans. Credit risk assessment nowadays is based on statistical or operational research methods. The statistical tools include discriminant analysis, which is essentially a linear regression, a variant of this is called logistic regression and classification trees, sometimes called recursive partitioning algorithm. The operational research techniques include variant of linear programming. Credit risk assessment has also lend itself to a number of different non-parametric statistical and AJ (Artificial Intelligence) modeling approaches. Once that have been piloted in the few years include the ubiquitous neural network, expert systems, genetic algorithm and nearest neighbour methods. In this work, we give an overview of the assessment of credit risk and used a combination of genetic algorithm and statistical approaches to overcome the limitations of a single technique for assessing credit risk as being "good" or "bad" with a view to trying to minimize the chance a customer will default on one particular lending to looking at how the firm can maximize the profit it can make from that customer. Statistral approaches to classification and discrimination in the field of risk portfolio was studied and a single function was arrived at, by merging the classification and discrimination function, and the function is then chosen as the fitness function for the genetic algorithm to carry out a selective search of higher order dependencies which converges to a solution within a reasonable time.