以共演化式遺傳演算法輔助動態股票投資決策分析

Using Co-Evolutionary Genetic Algorithm to Assist Dynamic Stock Investment Analysis

10.6846/TKU.2015.00635

夏承億

遺傳演算法 ； 共演化模式 ； 馬可夫決策 ； 動態股票投資決策 ； 最佳化 ； Genetic algorithms ； Co-evolutionary mode ； Markov decision process ； Dynamic stock investment decisions ； Optimization

淡江大學資訊管理學系碩士班學位論文

2015年

碩士

張應華

繁體中文

股票投資為時下眾多理財工具中的一種，其風險與報酬是相對的，想要獲得高額的報酬得承擔較高的風險，一般投資民眾缺乏專業知識與資料分析的能力，其投資決策多依電視股市交易分析節目或一些小道消息來進行決策，但股市的行情變化快速，投資者想要在股票市場裏獲利，必須慎選股票，並在適當的時機進場交易，以及最佳的資金配置。因此，如何制定正確的投資策略，成為眾多投資者關注的議題。 目前有許多決策方法可以輔助股票投資策略的制定，但是當投資者考慮的投資準則過多時，在判斷評估準則的重要性時容易失去客觀與正確的分析，尤其是當投資者臨時考慮到新準則時，傳統決策方法只能重新評量問題，無法累積之前所分析出準則的重要性，導致決策制定的不易，再者，投資人經常錯過股票買賣的時機與不知如何分配投資資金，有鑑於此，本研究利用共演化式遺傳演算法結合馬可夫決策過程來幫助股票投資策略的制定，利用遺傳演算法的平行空間廣度搜尋最佳解特性，加上模擬人類思考模式的共演化理論，使得遺傳演算法的執行過程可以隨著環境的改變而做動態的演化，從而找出最適當的進場時機、選股與資金配置，以搭配出一套完整的股票投資策略。

In the rapid changes stock market, investors want to get profit that they must carefully choose stocks, buy or sell the stocks at the appropriate time and with the best capital allocation strategy. How to make the right investment strategy is the subject of investors. There are many ways to assist the development of decision for stock investment strategy. But when investors consider too many investment criteria, it is easy to lose objectivity and proper analysis to determine the importance of the evaluation criteria. Especially when investors take into account new guidelines when to make investment decision. The traditional decision making methods just can only re-evaluate the decision problem and cannot be accumulated the importance of criteria before adding new guidelines. It is difficult to make decision. Furthermore, investors often miss the good opportunity to do stock transaction and do not know how to allocate investment funds. For this reason, this study integrated co-evolutionary genetic algorithm with Markov decision process to help investors develop a dynamic stock investment strategies system. The genetic algorithms have the ability of parallel search in breadth space. The co-evolutionary mechanism has simulated human thought patterns. These making the process of the genetic algorithm can implement with adjustment in the dynamic changes environments. And by Markov decision process, the decision system can inform investors when must to adjust the portfolio and to identify the appropriate stock timing, the stock selection and the capital allocation. The complete stock investment strategy allows investors to obtain excess returns when invest in the stock market.

1. 王聖方(2017)。整合隱藏式馬可夫模型與演化策略於股市交易策略最佳化。淡江大學資訊管理學系碩士班學位論文。2017。1-50。