Increase Debt Collection Using AI — Boost debt collection and recoveries using machine learning

This document describes an end-to-end real-world business use case that is successfully running in a production environment. As you may be aware, data is at the heart of every data science project. Because of the improper use of data elements, 85% of data science initiatives failed. The issue is a lack of knowledge about the data and how it should be used in the model. If the variables are properly constructed then fitting a model to the data is only 15% of the effort, and any developer can accomplish it by following the guidelines and model described in the document. The document describes all the data elements (variables) that were gathered and used in the model. How the attribute importance and dimensionality reduction were done. It describes the complete data pipeline with all the transformations that were applied. Furthermore, it explains step-by-step design in a way that can easily be understood and can be implemented by anyone who can program in R or Python. It also explains how historical data were labeled in order to build a training dataset for a classifier. It also describes how, with the aid of subject matter experts, several types of scoring models were developed that turned out to be extremely useful in a production environment. It also discusses why theoretical cross-validation of the model cannot guarantee results in a production setting, and why a different, more pragmatic cross-validation technique is used instead. It outlines which models were built and how they were created, as well as which data was utilized for model training and testing. It also specifies the benefits achieved when the model is run in production. This is not a tutorial for learning how to code. This paper is intended for data scientists that operate in a production environment and wish to build a successful machine learning model.