This project's goal is to forecast agent performance for the next quarter so that the company may make data-driven decisions to meet future sales targets. The company will be able to identify probable defaulters or overperformers and allocate additional resources accordingly.
I was working in collaboration with my team lead Yash Vedi on this project. In addition to development, there were a lot of back to back calls with clients on understanding the sort of data and their outcome expectation.
Some of the key development phases that were there are :
- ETL tasks with data from multiple sources
- Data exploration
- Building a timeline series of data
- Different scaler implementation for handling the outlier in data
- Model building with a lot of hyperparameter tuning
- Model Validation with out of time data
- Pipelines to push data to client system
- Reports for clients to view at Frontier Workbench