Estimation of domestic water demand using principal component analysis for densely populated area in Ajmer, Rajasthan

Estimation of accurate, domestic water demand is essential in water scarce urban and arid parts of western India cities for effective water resources planning and management. In this study, fifteen domestic water demand-variables are identified in a densed area of Ajmer City, Rajasthan (India). The collected data of fifteen independent-variables was used in the factor analysis of principal component analysis and new four PC (grouped variables) were formed. Based on these four PC variables, a multi linear regression called PCR (PC) best goodness-fit model was developed. This model (M4) gives the maximum value of regression coefficient R2 = 0.86. Later on, generalized least square PCR (GLS) and maximum likelihood PCR(MLH) techniques were also applied on these principal component group variables and two equations were proposed. The values of R2 = 0.59 and R2 = 0.60 were obtained for PCR (GLS) and PCR (MLH) respectively. Further, this newly proposed model PCR (PC) was compared with artificial neural network (Multilayer Perceptron) based on same four PC grouped variable whose regression coefficient was R2 = 0.87, which is very close to the value of regression coefficient obtained from PCR (PC) modelling. Henc proposed PCR (PC) based equation can be used effectively for estimating domestic water demand in densely area of city, Ajmer in Rajasthan state.