【作者】 叶翔；

【导师】 程从从；

【作者基本信息】 南昌大学 ， 计算机应用技术， 2012， 硕士

【摘要】 股票的趋势研究一直是股民关心的问题,研究的方法有很多。本文使用了数据挖掘的一个重要分支,关联规则方法来挖掘股票间的联动关系,统计3只股票带时序上涨的情况在过去的某个时间段中出现的次数。如果出现次数多,也就是支持度和置信度都较大,那么当“股票A第Ta天上涨且股票B第Tb天也上涨”的情况出现时,可以考虑在Tc天买入第三只股票。这里的Ta,Tb和Tc可取任意值。得出的规则可以用来辅助股票投资。本文的主要研究内容包括以下几个方面：(1)对国内关于股票方面的关联规则挖掘的相关文献进行分析和总结,对股票挖掘过程进行了深入探讨,对挖掘过程中的数据预处理、算法、关联规则兴趣度这三个方面已有的一些改进方法进行了概括和评价。(2)针对人们希望看到的“股票A在Ta当天上涨且股票B在Tb当天上涨,则股票C在Tc当天上涨,支持度是X%,置信度是Y%”这样的规则(Ta<Tb<Tc),提出了两种关联规则挖掘算法：算法一是基于时间窗口的关联规则挖掘算法,在股票时间序列上定义一个时间窗口,在时间窗口内循环查找2项集和3项集,通过时间窗口的移动寻找全部的2项集和3项集；算法二是bit-search算法,引入了比特串的概念,用比特串来表示连续的时间序列上股票的上涨信息。而比特串便于移位操作和逻辑运算,可较大的简化股票间的运算；也减少了需要的内存空间。根据结果的串的支持度计数是否满足最小支持度计数阈值,就能得到所需要的频繁2项集和3项集。(3)介绍了股票关联规则挖掘流程的设计以及对股票数据的预处理过程,进而将bit-search算法运用到股票数据挖掘中,并生成股票间联动规则。更多还原

【Abstract】 The researches of stock market behavior have always been issues of common concern to investors, and the research methods can be varied. This paper mainly used the association rules, an important branch of data mining, as a tool to mining the linkage between stocks, to count the occurrences of rising interval of three stocks statistically in the past within a certain time.If the occurrence rate was high, that is to say, the approval rate and confidence level was high, then when the situation as "A stock raises price on the date of Ta, B stock raises price on the date of Tb" occur, the investors could consider to buy a third stock on the date of Tc. The rule got by this way can be used to assist stock investment.The main content of this paper include the following aspects:Analyzing and summarizing the domestic literature which is about mining association rules on stock. The author conduct in-depth exploration into stock mining process, moreover the author conclude and evaluate some of the existing improved methods for data preprocessing, algorithm, association rules interestingness within the mining process.The investors would like to know the rules as "A stock raises price on the date of Ta, B stock raises price on the date of Tb. and from this we can get the information that C stock would raise price on the date of Tc.(Ta<Tb<Tc) The approval rate is X%, while the confidence level is Y%". Aimed at this, the author put forward two algorithm of mining association rules with time interval.Algorithm one is the algorithm of mining association rules based on time window. That is, applying a time window on the stock time series, and loop searching the two sets and three sets in time window, by moving the time window, find all of the two sets and three sets.Algorithm two is bit-search algorithm, introducing the concept of bit string to present the rising stock information within a continuous time series. As the bit strings is easy to carry on the shift operations and logic operations, it can greatly simplify the calculation of stock information and reduce the memory space needed. According to the result, whether the approval rate counts of strings meet the condition of minimum approval rate counts threshold or not, we can get the frequent two sets and three sets that we needed.(3) The design of stock association rule to mining model design and stock data preprocessing are introduced, and then the author puts bit-search algorithm into the application of stock data mining and conclude interaction rules between the stocks.更多还原