Research And Application On Time Series Prediction Based On Data Mining Method

Time series is recorded in accordance with the order of the time ordered. Since ancient times people have the needs of time series analysis and applications. In Ancient Egypt people calculated the time of agriculture production according to the time pattern of the fluctuation of Nile. As technology continues to evolve, the scientific level continues to improve, the field of research continues to be extended, the demand in Time series analysis of is increasing, in which the time series model fitting and prediction is one of the key people are very concerned. Time series data itself become high-dimensional, strong random direction, and the amount of data is increasing.Data mining technology as a combination of statistics, stochastic processes, computers, artificial intelligence, differential equations, and other disciplines, and involves the classical algorithm, genetic algorithm, neural network, wavelet analysis and other methods of technology, will be different It disciplines an effective means of dealing with massive data. Data mining for time series forecasting method is applied to more and more practice.The traditional method requires testing, new methods need to be introduced.This article mainly discusses about several aspects of research:1, We summed up several classic prediction models of Time series analysis, and analyzed and compared them.2, based on the former, we established a new method of predictive modeling, using the differential equations.3, we established a method of algorithm for the model to achieve the model.4, We fitted and predicted the financial time series data and the scientific observations time series data, respectively using the new model and the classical methods by SAS 9.2 software , and compared the results.