Churn Prediction

The Expresso Churn Prediction challenge is hosted by one of the biggest telecommunication companies in Senegal, Expresso, in order for them to figure out the likelihood for a customer to “churn”; that is, for a regular customer to stop using the company’s services(not making any transactions) after 90 days. Therefore, the goal of the competition is to develop a predictive model that would try to pre-emptively find the customers that will be more likely to become inactive according to the training data that the company provides. Compared to our other AI competitions, this one is a bit different in the fact that it’s a binary prediction task, which will rely heavily on the data side.

An example of the customer retention rates, found on here

One thing to note is that there’s a fair bit of data -in fact, 25 million entries- available to the teams, but the effectiveness of the algorithm lies in the preprocessing and cleaning up the data part. Compared to the other competitions, this one takes place without any baseline algorithms or any provided help by the hosts, and the only possible way to communicate with others is through the discussion board. However, this project is relatively easy to compute with modern equipment so the difficulty is leaning more on the algorithm side rather than hardware. Finally, for this competition, our team aims to use deep learning over machine learning to get better results.