A Prototypes-Embedded Genetic K-means Algorithm
ICPR '06 Proceedings of the 18th International Conference on Pattern Recognition - Volume 02
ICCIT '08 Proceedings of the 2008 Third International Conference on Convergence and Hybrid Information Technology - Volume 02
Forecasting time series with genetic fuzzy predictor ensemble
IEEE Transactions on Fuzzy Systems
Hi-index | 0.00 |
One of the most important goals of time series analysis is prediction basing on the analyzed information. But it is not easy to analyze the patterns, regularities and trends of non-stationary and/or chaos time series because their major characteristics are non-linear and vague. In this paper, we propose primary and secondary tuning procedures that can enhance the accuracy for designing fuzzy prediction systems. In the primary tuning procedure, the data preprocessing, model selection and general k-means clustering techniques are used to roughly tune the proposed fuzzy prediction systems. The primary tuning procedure is to choose the optimal difference candidates, partition the fuzzy sets for each candidate, and select the optimal difference interval (or predictor). In secondary tuning procedure, the real-coded genetic k-means algorithm (RCGKA) is used to enhance the efficiency of the clusters associated with non-stationary time series. The purpose of the secondary tuning procedure is to finely tune the fuzzy sets of the selected predictor. With two tuning procedures, the proposed prediction systems will reflect more clearly the characteristics of time series and predict more accurately the future values of the time series. Finally, in this paper, weverified the performances of the proposed prediction systems via typical time series simulations.