應用模糊分類元系統於股票技術分析

Abstract

分類元系統是Holland 及 Reitman 於1978年提出的，並成功的應用於事件回應問題上。所以 ,分類元系統非常適合處理部分情況未知、數學定義困難及動態環境下的問題。本研究應用模糊分類元系統學習股票型態，以做為股價預測模型。目的是希望設計能隨環境而產生或改變本身規則的股票預測學習系統，本研究以10日均線6個轉折點做為模糊分類元conditions，以第7個轉折點為系統action，並分析比較單及多模糊分類元系統執行結果，其中以多模糊分類元系統效果較佳。將來研究則可朝組合其他技術指標與成交量，比較其效能。

Learning classifier system is a special class of production systems first introduced by Holland and Reitman in 1978 and has been successfully used in a number of event-response problems. It is very suitable to describe a partially unknown environment and complex problems where it is very difficult to give a mathematical description and dynamic environment. In this research, we utilized the learning fuzzy classifier systems(LFCS) that learn the stock price patterns for forecasting stock price. The objective of this research is to design the stock forecasting model that is able to create and refine its rules in response to observe performance and changes in the dynamic environment. We use six turning points of 10 moving average to represent the conditions and seventh turning point of 10 moving average to represent the action, and the comparison with single LFCS model over multiple LFCS model that analysis the performance. The empirical evidence shows that MLFCS outperform more than that SLFCS .The further researches can be extended the model with other technical-indexs and trading volume.