In this paper, symbolic aggregate approximation which is the well-known dimensionality reduction for time series is utilized for enhancing previous approach to mine more useful group stock portfolio by grouping genetic algorithm. Each chromosome consists of three part that are grouping, stock, and stock portfolio parts. Grouping and stock parts represent how to divide stocks into groups. Stock portfolio part means purchased stocks and units. Each individual is evaluated by group balance, portfolio satisfaction and SAX distance. Experiments on a real data are conducted to show merits of the proposed approach.
Network-Based Information Systems (NBiS), 2015 18th International Conference, pp.280-285