Research And Application Of Feature Representation And Similarity Measurement Of Securities Time Series

The stock market is a barometer of the economy,China's stock market is huge,the total amount close to 40 trillion Yuan.The stock market is not only a good place for excellent enterprise financing but also brings dividend to investors,and it is the fortress of the national economic security.Researchers give more and more attention to the securities market,a large number of computer technology is introduced into the study of securities,a variety of techniques mostly depend on the study of a large amount of historical data.Stock trading produce a large amount of data every day,it is necessary to compress the data before the analysis is processed.Stock trading data has the characteristics of time series,The stock market theory knowledge should be fully considered in the feature representation of stock time series.In this way,we can keep the trend characteristics of stock time series as much as possible.In this paper,we present a new method of time series Feature Representation which based on the comparison of different feature representation methods and the thorough research of the stock theory.This method has the advantages of small fitting error,high compression ratio,easy to extend the data,and the trend of the stock time series is well preserved.The similarity measure of stock time series usually compare the price of the stock or amount of increase.This paper proposes a new similarity measure method,after the stock time series segmentation,according to morphological topology,Combine multiple segments into a trend.The trend of segment division will comply with the securities theory.With the trend of the various segments of the topology structure,to identify and distinguish the types of segments,and it's symbolic,using edit distance to measure the similarity of the stock.According to the shape of the stock to classify it,use the data mining function to find out the rules of change,as the basis for stock trading.