Prediction Of Time Series Data Based On Type-? Fuzzy Cognitive Map

At present,time series analysis technology is widely used in industry,business,research and other fields to support decision-making needs.In time series analysis,time series prediction is a very important research content.The main task of this work is to set up prediction model by analyzing historical data,and predict future value.Generally speaking,time series prediction models can be divided into two categories: linear prediction model and nonlinear prediction model.The linear prediction model is limited by its own structure and regression mechanism,and the prediction results of the time series data produced by the complex system are not accurate enough.The nonlinear prediction model,such as artificial neural network and other machine learning techniques,needs repeated training and iteration in the process of building the model,which reduces the efficiency of the model construction.In addition,both of them rely too much on sample data in modeling process,so that the models constructed are lack of "interpretability".Sometimes interpretability is a concern of decision makers.Fuzzy cognitive map is a knowledge representation graph.In its causal network,neurons and weights are precise for the problem domain.Therefore,the prediction model based on fuzzy cognitive map can be interpretable.However,the traditional prediction model based on fuzzy cognitive map(called Type-I model of fuzzy cognitive map)does not take into account the uncertainty of causality during the modeling process.In order to solve the above problems,the main research contents of this paper include:(1)summarize the development status of traditional time series prediction model;(2)study the theory of knowledge representation and causal reasoning of fuzzy cognitive map,and combine the two fuzzy set theory,and propose a model of Type-II fuzzy recognition graph(T2FCM),and The four layer neural network structure is used to construct the model.(3)finally,in order to verify the time series prediction ability of the T2 FCM model,the experiment is carried out on the two important time series prediction problems,which are the prediction of the MackeyGlass chaotic time series and the stock price prediction in the financial field.T2 FCM correspondingly compare to the prediction results of fuzzy time series model,ARIMA model and LSTM model.The experimental results show that the prediction model based on the Type-II fuzzy cognitive map has the ability to predict the Mackey-Glass chaotic time series and the stock price prediction in the financial field,and has obtained a better prediction result.