Application Of Data Mining In Stock Investment

In the past twenty years, data mining has drawn a lot of attentions from expertsand researchers in related areas and played an increasingly important role infinancial sector, telecommunications industry, retail industry, manufacturingindustry and health care sector. Securities industry early use information technologyto establish a relatively complete transaction processing system. After years of fastdevelopment, it has accumulated a large amount of data, within which a lot ofvaluable information is waiting to be explored, so how to transform those data intouseful information for securities investing has been the first priority of practitionersin the industry. Data mining’s abilities to deal with massive data and establishintelligent models are very meaningful to securities analysis. Therefore, the mainpurpose of this paper is to investigate applications of data mining technology in thestock investing field. The main work is as following:Starting with introducing data mining algorithms, we focus on the principlesand applications of data mining techniques, especially clustering analysis and NeuralNetwork Analysis. In order to solve the problem of when for investors to enter thestock market, neutral network models are established to predict the next day closingprice index. We not only consider the effect previous period data but also the historicdata may have on forecasting results, then the one-stage and multi-stage forecastingmodels are established to analyze the Shanghai securities composite indexes. Afterseveral experiments and comparison, a superior model is selected. Using that model,we predicted the next day close points of Shanghai securities indexes from Jan toApr,2010. By comparing the prediction results with actural values, we find thatthe established neutral network model has shown a satisfying performance.Furthermore, aiming at solving the problem of choosing individual shares forstock investors, this article will introduce the clustering algorithm. By selectingfinancial information indicators of individual shares and establishing an indicatorsystem, more than800A stocks are finally choosed for use. Clustering analysis isdone by Clementine software. Regarding to the model selection, TwoStep clusteringmethod is choosed to construct significant stock catogeries due to its betterperformance than K-means clustering and Kohonen clustering methods. Superiorstocks are finally distinguished for the effective decision makings of investors.