Analysis of the inclusion of quantitative methods for the improvement of the effectiveness of collection actions in a power utility

This work proposes the approaches of neural networks and logistic regression for selection of customers with electricity consumption bills late for the proposition of collection actions. Variables determined from business information such as debt amount, overdue invoices and time debt of residential customers of the Company of the State of Bahia Electricity (COELBA) are used in this study. The proposed methodologies were compared with each other and with the actual results, showing improvement in the overall efficiency of collection actions.