Application of Genetic Programming in Gold Price Forecasting

本文以2003至2009年的資料為研究區間，採用基本面分析指標、技術面分析指標及基因規畫法對倫敦黃金午後定盤價每季帄均塑造金價預測模型，同時歸納以基因規畫法塑造金價預測模型時，應使用何種投入指標與相關基因規畫法參數設定，較有機會獲得較佳預測力的金價預測模型。 最後發現對於黃金價格而言，各國股市大盤及黃金供需相關因素為使用基因規畫法塑造金價預測模型時較重要的指標種類，而於經濟狀況有劇烈變動時，加入技術分析指標將會改善模型的表現。而比較指標與基因規畫設定參數(如挑選函式、運算子集合、演化代數、染色體群大小)對模型預測力之影響，發現指標對模型預測力的影響遠大於基因規畫設定參數。The research uses the data between 2003 to 2009 to discuss the gold price forecastting model. Using fundamental analysis indices, technical analysis indices and Genetic Programming(GP) to modeling the gold price forecastting model. This paper also summarized that what kind of indexes and GP parameters should be set for getting better performance? Finally found that ,using the stock indices of important market and gold supply/demand factors to modeling usually get better performance. If there are drastic changes in economic conditions, using the technical analysis indices can improve the performance of model. The comparison of influence on model performance between indexes and GP parameters(ex. selecetio function, operator set, reproducting times, population size) show that, the indices have more influence to model performance than GP parameters