Research And Application Of Financial Time Series Forecasting Based On Data Ming

Won the Nobel Prize in economics in 1997, Robert Merton believe that the time value of money, asset pricing and risk management constitutes the cornerstone of modern financial theory, how in the uncertain environment for inter-temporal optimal allocation of resources is the most important problem. With the development of computer software, a variety of software and applications are widely used, and continued into the business, social production and life, etc. Finance is an extremely important part of social economy. On financial data for data mining, it is in a large number of transient, nonlinear and high signal-to-noise ratio of uncertain financial data mining valuable information.Data mining technology in the financial sector gradually rise. The traditional data mining techniques in the treatment of the common data performed well, but the processing of instability of financial time series to show some limitations. Thus improving existing data mining technology in the application research of financial time series is particularly important. In order to solve this problem, this paper based on the clustering in data mining research as the breakthrough point, the main work is as follows:First, in view of DBSCAN clustering algorithm cannot deal with the data sets of varied densities, combined with the initial point of optimization and parameter adaptive method for improving DBSCAN algorithm. This paper proposes a new data set can cope with change density spatial clustering algorithm based on density(OS-DBSCAN). The experimental results show that the new improved density based spatial clustering algorithm can deal with varied density data sets for clustering, and after giving the initial parameters according to the characteristics and attributes of data sets its own parameter adaptive, and compared with the traditional DBSCAN algorithm, the density of initial point optimization and parameter adaptive spatial clustering algorithm can improve the quality of clustering.Secondly, in view of the financial time series prediction based on data mining problems, combined with the proposed OS-DBSCAN clustering algorithm and SVR regression prediction algorithm based on particle swarm optimization, this paper proposes a hybrid algorithm to cope with unsteady, nonlinear and high signal-to-noise ratio of financial time series, and use the real financial data, such as Microsoft and Intel stock index data, TWII financial index data and 2259 A-share stocks trading history data, the practicability of the algorithm for further study. Experimental data shows that the hybrid algorithm of OS-DBSCAN and SVR based on particle swarm optimization in prediction of financial time series showed certain advantages, compared with the traditional way. The hybrid algorithm of OS-DBSCAN and SVR based on particle swarm optimization algorithm to improve the financial index regression prediction accuracy, can improve the predictive accuracy of shares on the next day. Data mining algorithm on the financial time series forecasting achieved certain effect and practicability.