Study Of The Time Series Forecasting Algorithm And System Realization Based On The ARIMA Model

In the domain of natural science and social science, the important policy-making question cannot be independent of the forecast because of the forecast is the policy-making foundation. Actually, the related thing's information is not completely, the related theory is also frequently imperfect, and people's understanding of things is limited to observation data, time series, it can only make use of the existing historical data structure model to predict the future.As an important research topic of data mining time series prediction developed rapidly in the last few years.Prediction is changing things foreseeable future development and Statistical data processing as the most effective means of the forecast to play an important role. Statistical projections is the use of statistical methods for the quantitative analysis of thing's that predict the outcome of a method or scientific. Many statistical forecast algorithms obtained the widespread application in the reality production life, for example: moving average algorithm, index smooth algorithm as well as ARMA method and so on. The statistics forecast technique establishment above the strict mathematical theory foundation, has the structure to be simple, forecast speed characteristics and so on quick, convenient operation, is opposite in other succession analysis forecast technique (for example: Return analytic method, nerve network method, grey theory and so on) more suitable practical application.In the present statistical predict that has the non-steady sequential analysis effect difference, many step prediction errors to be big, the deficient system's software to realize and so on questions. This article conducts the research in view of this question, proposed the NARIMA method, this method take the ARIMA model as a foundation, unified the tourist itinerary steady inspection procedure, the difference steady processing method, the linear smallest variance forecast algorithm effectively and so on, has solved the above problem which in the traditional statistics forecast technique exists.This article main innovation work is as follows:(1) In view of the bad characteristic of the non-steady sequence effect with the conventional routes, proposed the tourist itinerary inspection method. The main idea is : Carries on the supposition examination to the sample sequence, if the stability is dissatisfied, carries on difference processing to the sequence, the examination, satisfies until the stability. At the same time, the analysis result may trade the mapping through the contrast disintegration to the original sequence.(2) In view of the question that the many step prediction errors increase gradually of the tradition forecast algorithm, proposed the belt modification factor forecast algorithm.the main idea:induct error dynamic modification factor,Carries on the dynamic revision to the error which caused as a result of model structure's change in many step forecasting processes,increases the forecast precision greatly.(3) In view of each kind of analysis method's system research, present a new time series analysis method (NARIMA methods). The method includes: data preprocessing, model identification, parameter estimation, forecasting, analysis of the error. And present the multianalysis flow and the algorithm definition of this method.(4) In view of lacks system's software to realize the question ,developed a time series analysis forecast system platform, with the Java language development, and has used Swing, Struts, Hibernate, Spring and so on key technologies. This platform's algorithm realized further confirmed the NARIMA method convenience practical characteristic, simultaneously, has prepared for the following simulation testing.