Prediction of Commercial Insurance Payments for Surgical Procedure using DataRobot

Business Problem

• Predict the claim amount for MSA having less than 10 procedures.
• Validate the results for the DataRobot models developed by the J&J team.
• Develop a production level Machine Learning Framework for predicting Insurance Payments for surgical procedures.
• If possible, improve the model performance from the Baseline DataRobot models.
• Models and code to be delivered in Python.

Languages and Tools

Project Structure and Approach Outline

Documentation

Technical Report

Instructions for Project Setup

Please run the following jupyter notebooks in sequence:

1. Run the Final Code Snippet/Capstone_Step1_Data_Preprocessing.ipynb notebook responsible for transforming raw datasets into a preprocessed dataset. Please edit the file paths to data/raw folder.
2. Afterwards, run Final Code Snippet/Capstone_Step2_Data_Modelling.ipynbresponsible for performing data modelling and model evaluation.
3. Run Final Code Snippet/Capstone_Step3_XGBoost_Model.ipynbresponsible for XGBoost modelling and enforcing the monotonicity constraints. Please ensure that output of Step 1 and Step 2 has been generated before running Step 3.

Credits

Authors

• Rahulraj Singh - Team Captain [rs4211@columbia.edu]
• Prerit Jain [pj2383@columbia.edu]
• Mahesh Jindal [mj3038@columbia.edu]
• Parth Gupta [pg2677@columbia.edu]
• Ayush Baral [ab5247@columbia.edu]

Sponsors/Mentors (J&J Team)

• Cindy Tong - Project Lead and Domain Expert
• Ziyu Tan - Technical Expert and go-to person for DataRobot
• Katherine Etter - Mentor and Knowledge Expert

Instructor

Adam S. Kelleher [ak4063@columbia.edu]

CA

Xuanyu Li [xl3116@columbia.edu]