Utilizando machine learning para identificar clientes trials que se tornarão pagantes em um SAAS

Abstract

A SaaS system is a software delivered as a service in such a way that an user that wants to use it need only a way to access it, usually via a web browser. In this software delivery model, paid systems usually have an time range called trial in which an user can test the system before buying it. This issue has as a goal the identification of the users in trial who will become payers in a SaaS system. For that, the generated data by the customer’s system use are used in 3 classification algorithms: Support Vector Classifier, K Nearest Neighbours and Random Forest Classifier. Besides the user’s generated data, new statistical data generated in the execution of this issue showed to be useful on the customer’s subscription prediction. In the executed tests, Random Forest Classifier obtained better performance on the F1-score metric. In the end of the issue, a model is created to be used in the case company.