Prediction of watermain failure frequencies using multiple and Poisson regression

Abstract

An important concern for water utilities managers is the prediction of failure frequency of watermains. To provide insight, reliance can be structured based upon modeling of historical data. In this research two regression-based models are employed, namely multiple and Poisson regression models. The models are derived based on 10 years of historical data collected for the city of Sanandaj in Iran. Several tests to validate each of the models are described. The comparison of correlation coefficients for multiple and Poisson models, besides violating initial assumptions, show that multiple regression-based modeling is inadequate.