Medical Insurance Cost Prediction

A comprehensive machine learning project that predicts medical insurance costs based on personal characteristics using regression analysis.

📊 Project Overview

This project analyzes a medical insurance dataset to build a predictive model for insurance charges. It demonstrates a complete data science workflow from exploratory data analysis to model evaluation, making it an excellent learning resource for understanding regression analysis in healthcare applications.

🎯 Objectives

• Predict medical insurance costs based on personal characteristics
• Perform comprehensive exploratory data analysis (EDA)
• Build and evaluate a multiple linear regression model
• Understand the relationship between lifestyle factors and insurance costs

📋 Dataset

The dataset contains 1,338 insurance records with the following features:

Feature	Description	Type
age	Age of the insured person	Numerical
sex	Gender (male/female)	Categorical
bmi	Body Mass Index	Numerical
children	Number of dependents	Numerical
smoker	Smoking status (yes/no)	Categorical
region	Geographic region (northeast, northwest, southeast, southwest)	Categorical
charges	Insurance charges (target variable)	Numerical

🛠️ Technologies Used

• Python 3.x
• Jupyter Notebook
• Libraries:
  • pandas - Data manipulation and analysis
  • numpy - Numerical computations
  • matplotlib - Data visualization
  • seaborn - Statistical data visualization
  • scikit-learn - Machine learning algorithms
  • statsmodels - Statistical modeling

🚀 Getting Started

Prerequisites

pip install pandas numpy matplotlib seaborn scikit-learn statsmodels jupyter

Installation

1. Clone this repository:

git clone https://github.com/yourusername/medical-insurance-prediction.git
cd medical-insurance-prediction

2. Launch Jupyter Notebook:

jupyter notebook Medical_costs.ipynb

📈 Project Structure

medical-insurance-prediction/
│
├── README.md                 # Project documentation
├── Medical_costs.ipynb      # Main analysis notebook
├── insurance.csv           # Dataset
└── requirements.txt        # Dependencies

🔍 Analysis Workflow

1. Data Exploration

• Dataset Overview: Basic statistics and data types
• Missing Values: Comprehensive check for data quality
• Distribution Analysis: Understanding feature distributions
• Correlation Analysis: Identifying relationships between variables

2. Data Preprocessing

• Categorical Encoding: Converting categorical variables to numerical
• Feature Engineering: Creating dummy variables
• Data Transformation: Log transformation of skewed target variable
• Multicollinearity Check: Using Variance Inflation Factor (VIF)

3. Model Building

• Train-Test Split: 80/20 split for model validation
• Linear Regression: Implementation using scikit-learn
• Model Training: Fitting the model on training data

4. Model Evaluation

• Performance Metrics: R-squared score analysis
• Residual Analysis: Checking model assumptions
• Prediction Visualization: Scatter plots of predicted vs actual values
• Statistical Validation: OLS assumptions testing

📊 Key Findings

• Smoking Status: Strongest predictor of insurance costs
• Age: Positive correlation with insurance charges
• BMI: Higher BMI associated with increased costs
• Model Performance: Achieved test R² score of approximately 0.48
• The model achieved an R² of 0.76 on training data but only 0.48 on test data, indicating overfitting and the limitations of linear regression for highly variable medical cost data

🎨 Visualizations

The project includes comprehensive visualizations:

• Distribution plots for all numerical features
• Box plots for categorical vs numerical relationships
• Scatter plots for prediction accuracy assessment
• Residual plots for model validation

🔮 Model Performance

• Training R² Score: ~0.76
• Model Type: Multiple Linear Regression
• Key Insights: Smoking status is the most significant cost factor

Skills Demonstrated:

• Data cleaning and preprocessing
• Exploratory data analysis (EDA)
• Statistical analysis and hypothesis testing
• Machine learning model implementation
• Data visualization techniques
• Model evaluation and validation

🙏 Acknowledgments

• Dataset source: Kaggle Medical Insurance Dataset
• Inspiration from various data science tutorials and courses
• Open source community for excellent Python libraries

---

⭐ If you found this project helpful, please give it a star! ⭐