A predictive pipeline that evaluates the probability of a real estate tenant churn, and provides analysis of how different factors influence that probability.
Our client manages a lot of commercial real estate facilities in the US. Their aim was to reduce tenant churn. They could do so by making educated guesses about whether the tenants will churn and offering them appropriate discounts or other promotions and by understanding better how different factors influence tenant decisions.
- Integration with 3rd party cloud file manager
- Emails receivers and parsers
- Journaling service
- Scheduling service
- DB services for predictability algorithms
Using Machine Learning, we built a predictive pipeline that evaluates the probability of tenant churn and provides an analysis of how different features influence that probability. To get to this stage, we had to provide business analysis, figure out which features should this backend system have, have model training, and deploy the server.
So the project consisted of the following stages:
- Business evaluation
- Exploratory data analysis
- Selection and engineering of the features
- Model training and validation
- Machine Learning
The final result of the project is an ML pipeline that predicts probability of tenant churn, can continuously improve its accuracy by training on more data, and provides extensive model explainability. The system proved its efficiency by meeting customer’s historic data and being more accurate than it.