Mining of Ensemble Stock Timing Trading Rules Based on Gene Expression Programming

• Main Authors: Chen, DiHao, 陳帝豪
• Other Authors: Lin,WenShiu
• Format: Others
• Language: zh-TW
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/08657533238108543707

碩士 === 輔仁大學 === 資訊管理學系 === 100 === The main purpose of this study is to solve parameter design of traditional gene expression programming (GEP), and construct the optimal stock trading rules with ensemble learning strategy. Hope to solve the investment behavior of investor overconfidence and disposition effect. We explored the optimal parameter combination to resolve the problems of traditional GEP by integrating among GEP which is good at searching, encoding with combination of ensemble technical index and RNC, and Optimal Computing Budget Allocation (OCBA). Firstly, we constructed the relative return model which has the stable profit in GEP optimal stocks trading module. The relative return model was derived from nine ensemble technical indexes on the basis of Taiwan Capitalization Weighted Stock Index. Then, we conducted absolute return model with risk-free factor to evaluate the annual index return. The experimental results show that: (1) OCBA optimal parameters model yields the optimal parameter combination effectively. (2) The proposed ensemble stocks trading rules are more appropriate to short-dated training set and testing set due to a great number of discriminant indexes. (3) Both the proposed absolute return model and relative return model achieve good performances in training set and testing set. Especially, they yield the good return on investment in bear market. (4) The proposed module can apply in different performances stocks market in United States of America and China. In conclusion, the GEP-based ensemble timing trading rules can effectively resolve the problems of investor overconfidence and the irrational investing behaviors and increase the ability of investors in risk handling and investment performance.