Medical Insurance Cost Prediction

Predicted vs Actual Log Charges

Linear Regression model performance visualizing predicted vs. actual values on a log scale.

---

Project Overview

This project predicts individual medical insurance costs using demographic and health data. To improve model accuracy and handle the right-skewed distribution of insurance charges, a Log Transformation was applied to the target variable (charges).

Key Objectives:

• Analyze the impact of features like age, bmi, and smoker on total charges.
• Train a Linear Regression model using Scikit-Learn.
• Evaluate the model using standard regression metrics.

---

Model Performance

Based on the final evaluation, the model achieved the following results:

• R² Score: 0.894
• Mean Absolute Error (MAE): 0.21
• Mean Squared Error (MSE): 0.098

---

Project Workflow

1. Data Cleaning: Handled duplicates and verified no missing values existed.
2. Feature Engineering: * Encoded categorical variables (sex, smoker, region).
   • Applied np.log() to the charges column to normalize the distribution.
3. Training: Split the data into training and testing sets.
4. Prediction: Generated predictions on the log-scale and visualized them against actual values.

---

Tech Stack

• Language: Python
• Libraries: Pandas, NumPy, Matplotlib, Seaborn, Scikit-Learn

---

File Structure

• dataset/insurance.csv: Input dataset.
• images/linear_trend.png: Visualization of results.
• medical_insurance_cost_prediction.ipynb: Complete Python code and analysis.

---

Let's Connect!

Nosheen Khan on LinkedIn | Nosheen Khan on Kaggle