Microsoft-Forbes data analysis competition
In collaboration with Julia Petropavlovskaya
Predicting, whether a client is going to prolong the term of the contract or not
Metrics - AUC ROC
Data:
- Contracts
- Attendence
- Freezes
- Communication
All the data tables were merged into one, after cleaning, transforming and feature engineering.
Models:
- XGB
- Neaural Network
- Random forest
- Naive Bayes
- k-NN
- SVM
Models were tuned and optimized. Final prediction was made via ensebmling. Weighted average prediction method was used.
Final score on validation - 0.9846