A study on applying artificial intelligence and machine learning for modeling and predicting customer behaviors, churning and conversion

Abstract Digital companies have become an important provider of items, products, and services and they are increasingly replacing traditional markets. The growth of this business has created a heated competition among digital companies to extend their customer base and increase revenue. For this purpose, digital companies are now aware of the importance of gaining new customers and more importantly, maintaining existing customers as acquiring new customers is more expensive than retaining existing customers. That is why e-companies do their best to build strong bonds with their customers and support all efforts to predict possible churners and take proactive actions towards potential churners. In this paper we will build a framework based on time-series Markov model that performs both potential churn customers prediction and predicts visitors who tend to exit from the e-company without making purchases. Markov model is a statistical model able to observe states in temporal patterns of data. Proposed model will be implemented on public dataset called “RecSys2015” and we will compare its results with other algorithms for benchmarking.