Similar Mining Of Time Series In The Stock Time Series

Time series data is the data set that arranges every one according to the time, and it uses social, economic and technologic fields widely. It is not only the history record, but also the submerged and interesting mode contained. Since time series database is bigger and bigger today, it have important significance to how to analyze the huge data, for people to understand and make right decision. It's strongly necessary to study on data mining of time series.One of the important research directions of time series is to open out the internal laws of data by using data mining. Opposite to mature part of data mining (such as mining of database association rules and classify rules), mining of time series still falls into a new branch.Recently the study on data mining of time series mainly concentrates on both the similarity search and the pattern mining from a time series. Similarity search is just the research base of data mining on time series, since association, classify and clustering all need solve the similitude degree problem of time series. The main problem of the research in similarity search is that the time series data is too huge. One of the effective solutions is describing the data once again so as to reduce the data's number. And some of the description methods have been put out. The third chapter of paper combines the landmark model and PLR (piece-wise linear representation of time series), and bring forward a indicative method of time series piece-wise liner, base on key points (landmark according with some stated condition). The method distills the key points from original series, and fits each subsection liner fitting function by using maximum likelihood function and the method of least squares. The method combines the characteristics of two series' indicative methods. The paper puts out another computational method on similarity, which is not sensitive on many kinds of transfiguration of time series.Technologic analysis is a pop research subject all the while, and people put out all kinds of methods to forecast the trend of stock market. A classic component in the technologic analysis is the speciation analysis. Different speciation has different significance and implies some laws that could foretell the future trend of stock price. Finally, the paper discusses a method of technologic analysis on stock market base on data mining technology. The method is applied to fine the trend mode of stock market and forecast some variables such as stock index by similarity search of time series.The innovation of paper mainly manifest that combines the landmark model and subsection liner model, put key points (landmark that accord some conditions) as warranty, fit each subsection liner fitting function by using maximum likelihood function and the method of least squares. The excellence of the method is that it fulfils the experimentation output of physiology, considers the internal holistic characteristic of time series, also reserves the direction information of each subsection in time series, and filtrates the noise of actual time series. Another innovation of the paper is bringing the comparability mining technology of time series to the analysis of time series on stock market.