Predictive Model for Customer Churn

The project consisted in establishing an analysis of the real estate sector of a company, in order to identify who is most likely not to reserve a property and what type of property they are attracted to.

We worked on a database containing information about the type of property, the segment and the group of the property.

Information about the visitor, i.e. gender, group, occupation..., and information about the financial status of the visitor, such as income, household budget, personal income and monthly down payment.

This information allowed us to extract important information about the problem.

We then implemented a predictive model of the booking probability of these visitors, our goal was to maximize 'recall', i.e. the percentage of 'true positives correctly identified', as it is not true that people who do not book and identify themselves as bookers cost the company money.

Finally, based on the results of our EDA and our economic model, we were able to identify the most important and impactful variables in our business, which allowed us to propose short and long-term recommendations to address this issue.