Optimizing Diverse Group Stock Portfolio without Setting a Number of Groups

In the previous approach, by grouping genetic algorithm, an approach has been proposed for mining diverse group stock portfolio which can be used to generate various stock portfolios. However, a parameter, number of groups, should be given in advanced of that approach, and it is hard to set appropriate value. Hence, the intent of this research is to design an algorithm that can find appropriate number of groups automatically and a better diverse group stock portfolio from a given financial dataset. To allow exploration of different group sizes, the Davis-Bouldin index is considered to develop the fitness function for chromosome evaluation. In addition, the split and merge genetic operators are also utilized to increase and reduce the group sizes of a chromosome. Experiments on a real dataset were conducted to show the effectiveness of the proposed approach.