Study Of The Fitting And Clustering About Stock Transaction Data

With the rapid development of China’s stock market and the high degree ofstandardization of the market, stock varieties have a tendency to becomemulti-typical and multi-level. So stock market could attract more and more investors.In order to reduce investment risk and obtain high profit return, rational investorswill pay more attention to the choice of stock investments. Expressing real meaningof the stock data is critical for investors. Because of stock transaction datacontaining a lot of information, analysis of stock transaction data is particularlyimportant.The performance of the stock transaction data is affected by a number offactors, including a large amount of information. The stock transaction data reflectthe functional characteristics on the whole. There are many limitations to usetraditional time-series data analysis methods. According to this kind of the stocktransaction data, we analyses them by the functional data analysis method. The maincontent is preprocessing and curve fitting for stock transaction data based on thefunctional characteristics of these data. This method could make the original data"abstraction". So we can obtain a unified coefficient matrix, and then using thecoefficient matrix of reflecting stock characteristics function to cluster. In the end,we draw the corresponding conclusion about stock clustering, and these conclusionswere explained reasonably.Limitation of traditional analytical methods for functional data has beenimproved by using functional data analysis. This method not only increases thescope of the analysis of data, but also expands the applications of functional dataanalysis methods. In practically, fitting and cluster analysis about stock transactiondata by this new method could obtain the ideal result, which reflect the validity ofthe method. It can provide a better basis to investors for decision.