Classification of Basic Sanitation Services for Asset Valuation Using Random Forest

Abstract

In a regulatory agency's tariff review process, data consistency is of fundamental importance for better assertiveness. For this analysis, a large part of the highly relevant data is not informed, which leads to a manual process by the analysts responsible for the review. Aiming to assist the work, a case study was carried out with a qualitative and quantitative approach of the data aiming at extracting relevant information from a database made available with sewage and water supply assets, classification algorithms based on Machine Learning were implemented and validated. As a result, a Random Forest model capable of classifying the type of service in which the assets are inserted was developed, reaching an accuracy of approximately 80%. Thus, this work makes it possible to predict part of the missing information in reviews, which will reduce the agents' analysis time, in addition to reducing possible human errors in the process as a whole.