Abstract: This paper discusses problem of stock pattern dependency mining on data streams, proposes a mining algorithm. This algorithm includes pre-process steps such as time series segmentation and pattern presentation, using 4 item rule tuples to present the dependency and to calculate the confidence and support degree, and the synopsis structure for two and N data streams. Experiments on stock price data and a real system show that the algorithm is effective and can be used in forecasting.

Key words: data stream, data mining, pattern dependency mining

摘要： 针对数据流间“模式依赖”问题，给出了一种模式依赖挖掘算法，该算法包括：挖掘前时间序列分段和模式表示，条件规则元组的创建和维护，模式依赖的置信度和支持度计算，2个或N个数据流概要结构的设计等。股票数据实验和实际系统表明，该挖掘方法能够有效地发现数据流间的模式依赖，可用于预测。

关键词: 流数据, 数据挖掘, 模式依赖挖掘

CLC Number: 

• TP301.6