Predicting Users’ Responses of Public Utility Services - Multivariate and Neural Network Analysis - A Case Study

Abstract

This research addresses the problem of predicting the user’s responses through multivariate choice (MVC) and neural network (NN) frameworks for predicting quality, quantity and overall User satisfaction of public water supply organization, BWSSB (Bangalore Water Supply and Sewerage Board) in Bangalore - India for policy initiatives. The MVC study identifies statistically significant factors that explain users’ loyalty to express satisfaction and voice to express dissatisfaction. The MVC model correctly predicts 85% of satisfied customers across satisfaction dimensions. Wald test on 1940 responses confirms that there exits cross equation correlation across quality, quantity and overall Users’ satisfaction dimensions and thus appropriateness of MVC framework over traditional logit for predicting the user responses. NN framework outperforms the econometric model with 94% correct classification of user responses. The study opens up potential research opportunities for applying the advanced analytical frameworks for predicting user responses in various public and private settings for Policy initiatives so that the service providers could improve their service delivery.