| In real life, we know, stock price series move not only randomly, but also in a certain trend. Hence, forecast the future trend will have a very good guidance for investors to build a reasonable investment strategy. Different from the traditional fundamental analysis and sampling technique analysis, this paper proposed a method to predict stock yields behavior over a period of time in the future based on mining the pattern of transaction sequence, by this way, we can find the investment opportunity in future and make reasonable quantitative investment decision accordingly.Firstly, we defined the basic trend patterns of stock price series, and dug up the basic pattern libraries of stock price series. Secondly, we gave the definition of the distance between two transaction sequences. According to this definition of distance, we can measure similarity of transaction sequences in every pattern library. And then, by using cluster analysis method based on minimum distance, we found the basic patterns and compound patterns that crop up frequently, by this way, we dug up the pattern features that crop up frequently in the original stock price series, and then we extracted the characteristic value of pattern features. Matched the current sequence with the frequent patterns and judged the pattern feature of the current sequence. Finally, we predicted the follow-up yield behavior of current sequence over a period of time in future based on SVM classification techniques.This method predicted the possible investment opportunity in the future from the perspective of pattern features of stock price series. Then it has a strong practical significance, it provides a different angle to predict the possible investment opportunity. |
