Fuzzy Decision Tree Based Rule Extraction in Securities Analysis

While larger and larger pools of stock market data are available for investors, it is crucial for them to achieve the knowledge hidden behind and make the correct selections. The huge data amount, the variable data characteristic, and the noisy environment make this goal a great challenge. Using the model of fuzzy decision tree based rules extraction, a new set of fuzzy rules to select stocks with high returns is derived in this study. They are easy to comprehend and are proved to be efficient for picking these stocks. In performance evaluation, a portfolio is built for each year from year 1998 to year 2003 based on these rules. On average, the portfolio receives an annual return of 19.1%, much better than the average annual return, 5.8%, of the S&P 500 index.