An Intelligent Model for Generating Indications of Tax Gap in Service Companies

Tax gap is still one of the main problems of the Brazilian Tax Administration. In the case of the Service Tax (ISS), recognizing and estimating the tax loss becomes more difficult, as the ISS is a self-assessing tax and the services provided are volatile and cannot be verified after delivery. Aiming to promote tax self-regulation by service companies, this paper proposes a model for generating indications of tax gap, based on forecasting costs and arbitrating the billing of such companies. The model is composed by a committee of Artificial Intelligence (AI) and Data Science (CD) algorithms that infer a probability of a given company presenting outlier behavior. The differential of the model is the possibility of inferring such indications even in the absence of data on the costs of companies. The evaluation of the proposed model was carried out in a case study in the city of Fortaleza. As a result of the experiment, 1,839 service companies, contained in a universe of 22,071 companies, were recognized with strong indication of tax gap, resulting in loss of ISS tax revenue calculated at approximately R$ 10 million.