Hybrid Model for Water Demand Prediction based on Fuzzy Cognitive Maps and Artificial Neural Networks

Abstract

In this study, we propose a new hybrid approach for time series prediction based on the efficient capabilities of fuzzy cognitive maps (FCMs) with structure optimization algorithms and artificial neural networks (ANNs). The proposed structure optimization genetic algorithm (SOGA) for automatic construction of FCM is used for modeling complexity based on historical time series, and artificial neural networks (ANNs) which are used at the final process for making time series prediction. The suggested SOGA-FCM method is used for selecting the most important nodes (attributes) and interconnections among them which in the next stage are used as the input data to ANN used for time series prediction after training. The FCM with efficient learning algorithms and ANN have been already proved as sufficient methods for making time series forecasting. The performance of the proposed approach is presented through the analysis of real data of daily water demand and the corresponding prediction. The multivariate analysis of historical data is held for nine variables, season, month, day or week, holiday, mean and high temperature, rain average, touristic activity and water demand. The whole approach was implemented in an intelligent software tool initially deployed for FCM prediction. Through the experimental analysis, the usefulness of the new hybrid approach in water demand prediction is demonstrated, by calculating the mean absolute error (as one of the well known prediction measures). The results are promising for future work to this direction.