Abstract:
This thesis applies genetic programming to solve the investment problem of market timing and index tracking. Genetic programming is a computer algorithm used to find the optimal solution. The solution found by genetic programming is a local solution but it is recognized that the solution is close to the exact solution. In this study, genetic programming technique is used to find a profitable trading strategy so as to solve the problem of market timing. The trading strategy will generate both buy and sell signals based on the historical data of Stock Exchange of Thailand (SET) index. If genetic programming is able to find a profitable trading strategy, it means that investors could earn excess return over that of SET index when they apply the strategy with the index. However, in fact, investors cannot invest directly in SET index. So, they must form their index-tracking portfolio. Thus, this thesis will also apply genetic programming to solve for the combinations of index-tracking portfolio that generates daily returns as same as the index. This tracking portfolio is established from the securities having been the members of SET50 index. In this study, the tracking portfolio is consisted of 5, 10 and 15 companies. The combination of solutions for both market timing problem and index tracking problem is applied to the out of sample test in order to verify the possibility of the trading strategy becoming lucrative in the real practice. This thesis applies genetic programming technique to SET index data from year 1995 to 2003. The final results suggest that genetic programming dramatically help in finding both the profitable trading strategy and the tracking portfolio in training period. However, for out of sample period, the combination of the solutions found by genetic programming generates excess return over SET index with no strong evidence. Almost excess returns are positive but each t-statistic is not strongly significant. Hence, the study cannot conclude that the SET is the weak form of the efficient market hypothesis (EMH) in the study period.
