Discovering investment strategies in portfolio management: a genetic algorithm approach

Data mining on real stock data is performed using genetic algorithm. The basic idea is to use the relation of closing price moving averages of different lengths to guide the investment. By using the overall return rate to measure the performance of strategies over the training set, the problem of discovering investment strategies in portfolio management is converted to an optimization problem, which is solved by means of genetic algorithm Stock data from NASDAQ, including Microsoft, Intel, Oracle, and Dell are used for test purpose. Comparisons of genetic algorithm with random walk and exhaustive search are performed and results show evidence that GA is superior to these methods in term of the overall return rate for the test set.