Forecasting of Multiple-factor Fuzzy Time Series with a Steady-State Probabilities Model

Abstract

In this time of big data many people often using statistical or mathematical models to analyze historical data and use data analysis to predict the changes of future In past studies regression analysis neural network and logistic regression models or support vector machine etc are often used to analyze But the above methods are based on numerical data to accurately predict for the fuzzy data (semantic information such as weather information) these methods cannot be used to analysis Until 1965 Zadeh proposed the fuzzy theory the linguistic variable finally can be analyzed and discussion Currently the fuzzy theory has been widely used in many fields such as fuzzy time series fuzzy regression etc However there are some problems in the fuzzy time series forecasting models Like the limits of the numbers of variables ignoring the frequencies of fuzzy sets It make the large predictions error and the lack explanation ability Therefore in order to solve the above problems in this study we combine the fuzzy theory and the Markov theory using Markov matrix instead of the traditional fuzzy relation matrix consider the frequency of transfer status Calculate the steady-state probabilities and use correlation coefficient to combine each variable Build a whole new predict model and enhance the prediction accuracy of the results

Date of Award	2015 Jun 22
Original language	English
Supervisor	Sheng-Tun Li (Supervisor)