This project is a submission to Kaggle's Allstate Severity Claims competition (https://www.kaggle.com/c/allstate-claims-severity) on Kaggle. This was made as motivation for a Capstone Project for Udacity's Machine Learning Engineer Nanodegree.
The project consists of two main parts:
- EDA: The given dataset is thoroughly analyzed and visualized through various means.
- XGBoost Training, Testing, and Tuning: The powerful and classic XGBoost model is used in tandem with a series of hyper-parameter optimization techniques, including the reliable GridSearchCV, to produce an optimal model with which to predict the target variable, loss.