Genetic Algorithms for Financial Portfolio Selection
Abstract
In this thesis, we explore and classify the applications and improvements on GA in these researches, then design a conventional GA model which applies classic GA mechanism to select a portfolio from a 100 AMEX stock pool. Because the generations of this conventional GA model only evolves based on one fixed data segment, the results it yields are not stable enough for the extremely dynamic financial market. In order to keep the populations more dynamically reflecting the current market circumstance, we propose an adaptive GA model to improve the convention model. The enhanced model will only evolves certain number of generations using one segment of history stock data and change to another data segment. We implement both of the conventional and adaptive GA models and compare their performances using data of one same stock pool. The experimental results have shown that the adaptive GA model is more reliable and gains higher average Rate of Return. In addition, alternative GA operators are implemented and tested in the adaptive model to find the optimal solution to improve the Rate of Return.
Collections
- OSU Theses [15752]