Find in Library

Prediction of water consumption and its effective factors is an important step in water crisis management. Studies showed that meteorological parameters are considered as the most important factor for short-term prediction of water consumption. In this research, cluster-based sliced inverse regression method was used to identify the meteorological variables affecting the household water consumption in Qom. In addition to dimension reduction, this method can be used to remove collinearity. The data consisted of seven meteorological parameters and monthly household water consumption from 2001 to 2013. Data analysis indicated that instead of seven primary variables, only two new components which are linear combinations of independent variables can be used. The negatively charged maximum wind speed and relative humidity (0.757 and 0.4) of the first component, and the negatively charged average minimum temperature (0.753) and positively charged average air temperature (0.634) of the second component had the greatest impact on the components. The regression analysis indicated that the average minimum temperature coefficient 0.018, the maximum wind speed coefficient -0.004, and determination coefficient 0.92% are significant. Comparing the method proposed in this paper with the usual method of principal component analysis (PCA) for multivariate data analysis indicated that cluster-based sliced inverse regression has fewer errors. Moreover, noticing the impact of collinearity on the outputs of neural networks, the method proposed in this paper had better performance than the usual methods and consequently predicts water consumption.