Fuzzy Inference Based Autoregressors for Time Series Prediction Using Nonparametric Residual Variance Estimation
Federico Montesino Pouzols, Amaury Lendasse and Angel Barriga Barros
In: 17th IEEE International Conference on Fuzzy Systems (FUZZ-IEEE2008), IEEE World Congress on Computational Intelligence, June 2008, Hong Kong.
A new software tool for time series prediction by
means of fuzzy inference systems is reported. This tool, named
xftsp, implements a novel methodology for time series prediction
based on methods for automatic fuzzy systems identification
and supervised learning combined with statistical methods for
nonparametric residual variance estimation. xftsp is designed
as a tool integrated in the Xfuzzy development environment
for fuzzy systems. Experiments carried out on a number of
time series benchmarks show the advantages of xftsp in terms
of both accuracy and computational requirements as compared
against Least-Squared Support Vector Machines, an established
technique in the field of time series prediction.