Models for adequately estimating water consumption in Taiwanese government institutions were developed to assist the government to more accurately predict and account for their water needs. A correlation coefficient matrix of associated factors was constructed based on records per unit of water consumption, describing the impact of various water consumption factors. To understand and quantify the effect of the impact factors, linear and nonlinear regression models, as well as an artificial neural network model were adopted. To account for data variability, the data used for modelling were either fully or partially adopted. For partial adoption, the quartile method was employed to remove any outliers. Analysis of the factors affecting water consumption revealed that the building floor area and number of personnel in an organization had the largest impact on estimated consumption, followed by the number of residential personnel. As the coefficient of variation for the green irrigated area and number of consulting personnel was low, the total area and the total number personnel of water consumption decreased the effectiveness of the model. Full article