基于最小首差循环链码的快速搜索算法  

Fast Search Algorithm Based on the Minimum First Difference Chain Code

作　　者：曹茸[1] CAO Rong （School of Information Engineering, Eurasia College, Xi＇an,710065, China）

机构地区：[1]西安欧亚学院信息工程学院,陕西西安710065

出　　处：《电脑知识与技术》2009年第7期5138-5139,5147,共3页Computer Knowledge and Technology

摘　　要：分析了常用的时间序列相似性搜索的一般方法，提出了基于最小首差循环链码的快速搜索算法。其基本思想是对股票数据进行合理的符号化，并对符号化后的数据求最小首差循环链码，有效的解决了因时间轴上的平移、伸缩、旋转等对相似性判断的影响，遇到相似序列后跳到该相似序列的下一个位置继续搜索，大大提高了相似性搜索的效率和准确率．并给出了理论分析和验证结果。On the basis of analyzing the frequently-used time-series similarity search algorithm, a kind of fast search algorithm based on the minimum first difference chain code, is presented, which first symbolizes stock data reasonably, then gets the minimum first difference chain code based on the data symbolized, which avoids the influence of time translation, expansion and rotation to similarity judgment effectively, and goes on to search in the next position when coming across similar serial, as well greatly enhances the efficiency and accuracy of similarity search, followed by theoretical analysis and validated results.

关 键 词：数据挖掘 时间序列 相似性搜索 子序列匹配 最小首差循环链码 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构;自动化与计算机技术—计算机科学与技术]