Forecasting-Mining Prediction of Water Consumption for Residential Sectors

Abstract

Forecasting of water consumption is done to determine the management decisions and choices of investments for urban water management. The data is of a single family’s the annual water consumption. The data is cleared of the unusual fluctuations on few days due to external influences using the sampling methods. The various methods used for forecasting the future consumption are, the fuzzy c-means prediction, the BIRCH (Balanced Iterative Reducing and Clustering using Hierarchies) prediction model and mining - forecasting model. The comparison is based on the correctness of its prediction against the following years. The results highlight that using mining - forecasting techniques predict a more accurate solution compared to the other methods. The mining – forecasting method causes the least deviation of 5%, while the fuzzy c-means predicts with a 58% deviation and BIRCH causes 38%.