The Time Series Forecasting Model Based On Artificial Intelligence Integration Technique And Its Application

The research of time series prediction,which has the vital significance in the field ofdecision making and risk management,is a particularly important content of the prediction domain.The traditional time series analysis models such as ARIMA model are usually based on assumption of normal distribution and unable to capture the nonlinear characteristics of time series data.So for most practical problems,the traditional models are no longer applicable.Artificial intelligence(AI)technologies like ANN and SVM overcome the limitations of the traditional model and fully present the nonlinear dynamic system of time series.The widely use of AI in the field of time series analysis has proved its good performance and universality.Moreover,many theoretical and empirical studies have also prove the authenticity and effectiveness of the integrated thought.Yet the improvement of existing linear integrated technology is very limited.The rise of the study of nonlinear integration technology tries to utilize AI to study weight pattern,which immensely ameliorates the fitting precision and generalization ability.This article introduces three kinds of time series prediction model,namely the S-BPNN,the EMD-LSSVM and nonlinear integrated model combined with the statistical theory,AI and integration technology,and empirically analyzes the spot price data of the Europe's Brent Crude.The paper compares the prediction performance of all models based on ARMA-GARCH model according to the normalized mean square error and the statistics of direction change.The results of empirical analysis indicate that: the prediction performance of EMD-LSSVM model is the best among three separate time series forecasting models from perspective of both performance evaluation indexes;The overall performance of integrated prediction model is superior to the single model;In the three integrated forecast model,the forecasting performance of the nonlinear integrated model based on SVM is better than others.That's to say,the nonlinear integrated model based on SVM tends to achieve the most accurate prediction and then provide the more favorable investment and policy recommendations to investors.