Predicting customer satisfaction for distribution companies using machine learning

Abstract

Purpose

This study aims to support electricity distribution companies on measuring and predicting customer satisfaction.

Design/methodology/approach

The developed methodology selects and applies machine learning techniques such as decision trees, support vector machines and ensemble learning to predict customer satisfaction from service data, power outage data and reliability indices.

Findings

The results on the predicted main indicator diverged only by 1.36 per cent of the results obtained by the survey with company customers.

Research limitations/implications

Social, economic and political conjunctures of the regional and national scenario can influence the indicators beyond the input variables considered in this paper.

Practical implications

Currently, the actions taken to increase customer satisfaction are based on the track record of a yearly survey; therefore, the methodology may assist in identifying disturbances on customer satisfaction, enabling decision-making to deal with it in a timely manner.

Originality/value

Development of an intelligent algorithm that can improve its performance with time. Understanding customer satisfaction may improve companies’ performance.

Citation

Cavalcante Siebert, L., Bianchi Filho, J.F., Silva Júnior, E.J.d., Kazumi Yamakawa, E. and Catapan, A. (2021), "Predicting customer satisfaction for distribution companies using machine learning", International Journal of Energy Sector Management, Vol. 15 No. 4, pp. 743-764. https://doi.org/10.1108/IJESM-10-2018-0007