Churn Prediction Estimation Based on Machine Learning Methods

Abstract

Customer churn prediction is classic task of machine learning, the relevance of which continues to grow. This is due to the fact that business companies collect more data about their customers and their behavior every year. A model that predicts whether a customer churn will occur in the future allows a business to build an optimal personalized pricing policy to retain a customer. Existing approaches for solving the problem of churn prediction for different areas are analyzed. A strategy for determining the period of customer churn is proposed and the optimal variant of data labeling is selected, which allows to convert the problem to a typical binary classification problem. A set of data from the Prozorro system was chosen for practical application of the approaches. Ensemble tree methods (Random Forest, XGBoost, LightGBM) were chosen as learning algorithms. Customer churn prediction is one of the many applications in today's world. Knowledge of mathematics and machine learning algorithms, along with the correct ability to build a problem statement - are the key skills of a specialist who studies it.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 7 citations exist upstream.