Allstate_Claims_Severity_Competition

Allstate is currently developing automated methods of predicting the cost, and hence severity, of claims, which is a typical regression problem. One interesting thing about this competition is that the optimized objective is MAE (Mean Absolute Error). Finally, I ranked top 11% in this Kaggle Competition.

Data Profile

Data Type	Number
Training Data Size	188318
Test Data Size	125546
Number of Features	130
Number of Continuous Variables	14
Number of Categorical Variables	116

Feature Engineering

• Categorical Feature Interaction
• Categorical Feature Alphabetical Encoding
• Numeric Feature Unskewness
• Numeric Feature Normalization
• Log + Shift Transformation of Target Value

1st Level Models

• Xgboost
• Neural Network
• Random Forest
• Extra Trees
• Regularized Greedy Forest

2nd level Models

• Stacking (Xgboost)
• Weighted Average

Tools

• Numpy
• Pandas
• Scipy
• Sklearn
• Xgboost
• Keras
• Bayesian Optimzation: https://github.com/fmfn/BayesianOptimization/blob/master/examples/xgboost_example.py