Research And Implementation Of Stock Informatiom Processing System

The stock price series is a typical time series. It has a great theoreticalsignificance and practical significance by utilizing time series techniques to researchand analysis the stock price series. According to the research on the theory andmethod of time series, and applying to the stock price series, it can find the internalregularity of stock time series and analyses the stock’s potential trend. The goal of thispaper is to utilize stock time series modeling technology and provide the investors anautomatic and intelligent stock analysis tool. The main content of this paper asfollows:(1) Design and implementation of a stock information processing system, whichinclude user management, technical analysis, stock condition and other basicfunctions of stock analysis.(2) Proposing an improved time series fitting algorithm. The main idea of thealgorithm is to find the stock key trends point by combining the slope method andtriangular midline method, then piecewising linear fitting the time series. Last wemake the implementation of the fitting algorithm in the stock system. Theexperimental results show that the improved algorithm of time series representationhas a good effect on data compression and the stock trend extraction. Finally, puttingthe improved algorithm into the stock information processing system, it shows theadvantages in operation conveniently and obtains the key information rapidly.(3) Pointing at the problem how to choose inputting feature vector to make thesupport vector machine forecast more precisely. This paper proposes two improved algorithm: the one is utilizing the key piont algorithm to select the feature of theoriginal stock information, selecting the feature which can represent the whole stockseries in the stock series as the input feature vector of SVM; The other is according tothe information gain method of the decision tree to determine the importance of inputstock features, then the weighted feature as the SVM’s input feature vector by theinformation gain values. Finally, putting the two improved algorithm into the stockinformation analysis system and it improves the precision of prediction greatly.In conclusion, firstly this paper implements the basic functions of stock analysis.Additionally, with the combination of time series algorithm metioned above thissystem implements two unique functionalities which are different from other stocksystems. The first one is SVM predict method based on key point. And the other oneis SVM predict method based on the feature selection of the decision tree.