The Investigation And Application Of Data Mining In Time Series

Data mining has been considered as a important methods in the analysis of time series, received more and more attention and research came from scientists. It is necessary and challenging subject to research how discovery valuable knowledge in large-scale time series database, and how to search based similarity while user give a graphic query pattern.This paper has investigate the similarity of time series and the modeling of construct a decision tree, which is based on the analysis of characteristic and practically applied requirement of time series. The main work and innovations are as follows:1) Deeply and widely analysis on the basic theory.This paper has arranged couples of basic theory in many chapters, include data mining, time series, Decision tree and PSO algorithm. And it has deeply and widely analysis on those basic theory which is a nice introduction for the later chapters.2) Investigation about the similarity search of time series.The third part of this article has proposed a method for dimension reduction and similarity search based on regression coefficient which has simplified the original series in utilizing the trend of time series. Then it has researched the feasibility of similarity search when using correlate analysis, and has given a empirical research in the end whose conclusion shows there is a highly practical meaning of the above work.3) The modeling of construct a decision tree using advanced PSO algorithmThe fourth past of this article has proposed a new model of construct a decision tree using advanced PSO algorithm. The model has fused the information augmentation ratio used in C4.5 algorithm and advanced PSO algorithm to build a decision tree. The final experiment decision illuminate the model has some advantages on the speed and accuracy of decision tree construction when compare with C4.5 algorithm.