Mining Multivariate Heterogeneous Time Series Models with Computational Intelligence Techniques

J.J. Valdés and A.J. Barton (Canada)

Keywords

Neurofuzzygenetic Architectures, Parallel Data Mining, Signal Processing, Neural Networks, Genetic Algorithms, Forecasting and Prediction

Abstract

This paper presents experimental results of a computational intelligence algorithm for model discovery and data mining in heterogeneous, multivariate time series, possibly with missing values. It uses a hybrid neuro-fuzzy network with two different types of neurons trained with a non-traditional procedure. Models describing the multivariate time de pendencies represent dependency patterns within the sig nals and are encoded as binary strings representing neural networks (evolved using genetic algorithms). The present paper studies its properties from an experimental point of view focussing on: i) the influence of missing values, ii) the factors controlling the model search process, and iii) the effectiveness of the time series prediction results. Re sults confirm that the algorithm i) possesses high tolerance to missing data, ii) has an error distribution skewed towards lower error values, iii) is capable of learning good models within large signal sets.

Important Links:



Go Back