Using Evolutionary Computation to Predict the U.S. Dow Jones Index

• Main Authors: CHENG TUNG HUI, 鄭東輝
• Other Authors: 林維垣
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/04618621052696688965

碩士 === 東吳大學 === 經濟學系 === 101 === Abstract In addition to the application of Genetic Programming (GP) and traditional multiple linear regression model, the purpose of this study is trying to use the new method of artificial intelligence –Artificial Bee Colony Algorithm (ABCA) to optimize the generalized regression neural network (GRNN) parameter, and to predict Dow Jones stock Index change, for academic research references. Experimental design is set as follow: the closing price of weekly data of Dow Jones index from 1996 to 2011 are collected for empirical analysis. The empirical results show: (1) the performance of artificial intelligence methods (GRNN, GP, and ABCA) to build predictive models were significantly better than the traditional linear regression. (2) GP prediction model established is more effective than GRNN. (3) The use of collective wisdom of the colony algorithm to optimize the GRNN smoothing parameter would improve its performance (default value of GRNN). (4) Of intelligence methods, ABCA is superior to GP and GRNN to predict Dow Jones stock Index change. Therefore, we deeply believe that the group wisdom algorithms (including ABCA) will play active roles in the artificial intelligence, and share valuable contributions in the future.