Svm-based Temporal Data Mining And Securities Analysis In Research

Prediction and periodicity detection are regarded as two of the important fields intemporal data mining. Prediction is the method of constructing a temporal miningmodel to forecast the unknown data in the future with the data we have known in thepast or at present. Periodicty detection is to find the repeat pattern of temporavl data.After support vector machine (SVM) was proposed in middle of 1990s, it is a newtool of data mining now, which has the advantages of global optimization, simplestructure and high practicability. By applying SVM to temporal data predictionmining, this paper will solve the classification and regression problems, such asprediction issues in the fields of finance, meteorology, hydrology, supermarket andmedicine. This paper has specific theoretical significance and practical value.This paper consists of five chapters, and the main work is as follows.Chapter 1 describes the researching background and relative technology oftemporal data mining, analyzes researching situation, discusses SVM and itsapplications in temporal data prediction mining, and sums up the results andinadequacies in this area. The research significance and content of this paper areintroduced.Chapter 2 studies the relative concepts and theories of temporal data. Temporaldata prediction model and periodicity pattem are built based on the relative definitionsand properties of it. This chapter provides the theoretical principle to the next study.Chapter 3 presents the theory of SVM, on the basis of statistical learning theory,introduces support vector classification (SVC) and support vector regression (SVR)respectively, and discusses the relationship of them.Chapter 4 gives the method of constructing the temporal data with the primarytemporal data, proposes the way of temporal data prediction based on the improvedsupport vector machine regression,ν-SVR.Chapter 5 expands the model of temporal data, introduces the temporal dataperiod pattern, takes advantage ofν-SVC to build the method of recognizing the temporal data period pattern.In this paper, we obtain the following results.Firstly, we construct the temporal data prediction model based on SVM.Secondly, method of temporal data period pattern recognition is presented in thispaper. Finally, we propose the way of constructing the temporal data using theprimary temporal data.