Applying Clustering Algorithms to Construct a Stock Trend Decision Model

Abstract

The stock market is a well-developed and mature market. Nevertheless, it is not immune to international financial market changes, where volatility has reigned in recent years. Investors who misgauge stock trends can suffer dramatic losses. Accurate identification of market trends can still achieve outstanding performance and has become a major investor concern. This paper proposes a new stock price trend clustering model using a genetic algorithm to search for optimal investment strategies. Daily stock prices and trading volume data from the Taiwan stock exchange weighted index (TAIEX) was used to examine the proposed trend clustering model’s performance. The model was also compared to other popular stock market investment strategies to verify its validity. Research results confirmed that the trend clustering model correctly identified three different trends in the stock market. Furthermore, the trend investment strategy model using genetic algorithms performed better than other investment strategies, i.e. Granville’s rules for buy and hold strategies, in both bull and bear markets. Research results confirmed trend investing outperformed the other two investment strategies in return and capital distribution, both during the training period and the testing period.