AV-Dataverse-Hack---Insurance-Claim-Prediction

Competition hosted on Analyticsvidhya

About

Create a machine learning model to predict if the policyholder will file a claim in the next 6 months or not based on the set of car and policy features.

Final Competition score is 0.1749408983

Leaderboard Rank is 38

Evaluation Metric is F1-score.

File information

• av-dataverse-hack-insurance-claim-prediction-eda.ipynb

  Basic Exploratory Data Analysis

  Packages Used,

   * seaborn
   * Pandas
   * Numpy
   * Matplotlib
  

• av-dataverse-hack-insurance-claim-prediction-model.ipynb

  Data Pre-processing and model.

  Packages Used,

    * Sklearn
    * Pandas
    * Numpy
    * Matplotlib
    * shap
  

  Created Random forest classifier model and evaluated with f1-score.

  For more detailed information about the model.

Random Forest Model Feature Importances

Random Forest Model - SHAP Feature Importances

SHAP Top feature influences the class 0

SHAP Top feature influences the class 1

Threshold Tuning Results - Optimal threshold: 0.0745, F1-score: 0.16929