Life Insurance Risk Prediction

Deciding for Life Insurance for a person using machine learning based on a his/her health & history that includes BMI, Age, Ht, Wt, Employment_History, Insurance_History, Medical_History etc.

Dataset

The Prudential Life Insurance Assessment dataset has been used in this project. This dataset consists of over a hundred variables describing attributes of life insurance applicants and over 50k rows (no. of applications).

Exploratory Data Analysis

For obvious reasons, BMI is the one of the most important factor for deciding the risk.	Same for the age of person, more preferable option for insurance are young people.

Clearly shows most of people who got insurance had BMI between approx. 0.12 and 0.50 Rest features doesn't show any patterns.

ML Model

I used Random Forest Classifier model for prediction. Used GridSearchCV for hyperparameter tuning with roc_auc as scoring metric. Acheved an accuracy of 80.8% in training data and 80.4% on test data.

Confusion matrix for training data	Confusion matrix for test data

Model Interpretation

Important features for prediction
Understanding test model prediction using SHAP