An Investment Study Of High Stock Dividends And Stock Splits Based On Pattern Recognition

“High stock dividends and stock splits” is the abbreviation of bonus issue or accumulation fund turning with high proportion, which is a market event that occurs frequently and has significant positive excess returns in a period of time before the plan announcement day. Therefore, the investors can get a better return on “high stock dividends and stock splits” investment if they recognize the event before announcement. However, the research literature on the "high stock dividends and stock splits" focuses on the dividend policy and wealth effect verification. While many domestic securities investment institutions have research on "high stock dividends and stock splits" investments, but most of the institutions use the method of scoring sort which is more subjective, and so that this paper presents a more objective pattern recognition forecasting model.Firstly, this paper defines the event connotation and mode of "high stock dividends and stock splits", and extracts characteristic factors of its listed Companies which contain share capital reserves, retained earnings per share, EPS, net assets per share, net cash flow per share, revenue share, the length of time to market, stock prices and capital stock. On this basis, the paper selects the method based on principal component logistic regression to establish a probability prediction model about Whether a stock implements the "high stock dividends and stock splits" event, and we build a predictive model with the data of 2014 to predict stocks that may implement "high stock dividends and stock splits" in 2015 after the model having been tested and co-linear processing. The forecast result is that the prediction accuracy rate of 2015 is up to 62.8% far higher than the 19.67% of accounting on A shares. Then the paper builds a simulation trading strategy by using the CART decision tree model to filtering the previous prediction samples. The annualized rate of return is as high as 122.41% better than the trading strategy which is unfiltered.Significance of this study mainly exists in the following two aspects. On the one hand, the paper build a complete portfolio strategy about "High stock dividends and stock splits" from an investor’s point of view, which has a certain reference value to investors. On the other hand, the use of data mining methods to study the "high High stock dividends and stock splits" investment provides new research ideas for future research.