Mining group stock portfolio by using grouping genetic algorithms

In this paper, a grouping genetic algorithm based approach is proposed for dividing stocks into groups and mining a set of stock portfolios, namely group stock portfolio. Each chromosome consists of three parts. Grouping and stock parts are used to indicate how to divide stocks into groups. Stock portfolio part is used to represent the purchased stocks and their purchased units. The fitness of each chromosome is evaluated by the group balance and the portfolio satisfaction. The group balance is utilized to make the groups represented by the chromosome have as similar number of stocks as possible. The portfolio satisfaction is used to evaluate the goodness of profits and satisfaction of investor´s requests of all possible portfolio combinations that can generate from a chromosome. Experiments on a real data were also made to show the effectiveness of the proposed approach.