Research on extracting excess earning stocks with fuzzy association rules

In this paper, we introduce a new approach to acquire excess earning portfolios at stock market We treat price, volume, price's swing, etc. as attributes of database and form segments by sliding a window of width W through record-sets in the database. The subsequences are clustered into k basic pattern with fuzzy clustering method and the proportion of the numbers exceeding the index's earning to the numbers of subsequences in a basic pattern is considered as the confidence of the pattern having excess earning. Decision tree and information gain are used as a heuristic for selecting the attribute that will best separate the segments into sub-segments. Stocks with maximal confidence are chosen as portfolios.