Dominion Akinrotimi
Back to Projects
Python scikit-learn Regression Model evaluation

Insurance Premium Prediction: Proving a Dataset Has No Signal

The brief: build a model that predicts insurance premiums for SecureLife Insurance Co., the fictional company in a 3MTT capstone, from 278,860 synthetic policy records with 20 columns.

What I found: none of the 18 customer and policy features predicts the premium. Four models, including two tree ensembles, scored an R² of about zero on 55,404 held-out policies. They do no better than quoting every customer the average premium. Showing that clearly, and saying what data would work instead, is the result.

View the notebook on GitHub

Results

ModelMAERMSER²
Ridge Regression695.0897.2-0.0002
Linear Regression695.0897.2-0.0002
Gradient Boosting695.0897.3-0.0003
Random Forest719.9908.5-0.0255
Ridge, tuned (alpha 100)695.0897.2-0.000

The premium's standard deviation is about 895, so an RMSE of 897 is what a constant guess scores.

Why I trust it

Correlation heatmap where every feature has near-zero correlation with Premium Amount

The strongest correlation with the premium is Credit Score at 0.005. Every other feature is closer to zero.

Scatter plot of premium against credit score showing an even cloud with no trend

Premiums spread evenly from 0 to the cap at every credit score. Income, health score, previous claims and age look the same.

A correction I made: my first report ranked the Ridge coefficients as "drivers" of premiums (education, exercise, smoking). The largest is about ±9.5 on a premium that varies by about 895, so those values are noise. I withdrew that reading rather than present it as insight.

What the business would need

What I built

Have data you cannot make sense of yet?

I turn raw data into dashboards, reports and decisions your team can act on.