The Application Research Of Time Series Data Mining On Stock Forecasting And Analysis

With the rapid development of information technology, the ability of generating and collecting data by information technology has improved greatly. How to solve the problem "data rich and information poor", so data mining comes into being. As the important area of data mining research, time-series data mining and forecasting have developed greatly. While as the particularity of data description, researchers pay much attention to how to apply the traditional data mining technologies to time-series data mining and forecasting.Based on the latest research on time series data mining, this dissertation researches on time series analysis, time series representation and measuring, similarity search, piecewise linear representation of time series, etc. Improved algorithms have been proposed in segmenting time series and similarity search, and considerable achievements have been made in the final analysis of the stock forecast a prototype system. The detailed work of this paper is as follows:①The time series analysis methods and characteristics was studied. Discussed how to choose the time series models. It investigates the time series data mining applications.②It discussed the transformation and representation of the time series, and those methods were compared. The time series similarity measure and its characteristics were studied. Did research on the segmenting time series and its algorithm were discussed in sliding window method and clustering analysis. A novel algorithm based on clustering to improve the accuracy of the segmenting time series was proposed. Did simulation for financial data from which the results are analyzed and discussed.③It investigated the time series similarity search. It discusses the time series data of indexing technologies and their merits and faults. The time series representation KL method based on the important sub-point is studied. For the special characteristic of financial data, based on the current searching methods with similar subsequence, a method called MABI(moving average based indexing) is proposed to effectively deal with the issue ofε-search query in subsequence matching. This method can be quickly eliminated a large proportion of non-qualified candidates, has greatly narrowed the scope of the search, laboratory tests show that the method than the traditional method has obvious advantages. ④In the comparative analysis of time series analysis of the existing features of the tool, put forward an integrated time-series data mining prototype system model framework to achieve the time series data mining, similarity search and analysis functions of the stock forecasting.