marta.github.io

Insurance Costs Analysis

This project explores the “U.S. Medical Insurance dataset” to understand which factors influence individual insurance pricing. The analysis includes data exploration, visualization, and a multiple linear regression model using statsmodels.

---

📊 Dataset

The dataset contains 1,338 observations with the following variables:

• age: Age of the individual
• sex: Male / Female
• bmi: Body mass index
• children: Number of children/dependents
• smoker: Yes / No
• region: Residential region in the U.S.
• charges: Medical costs billed

---

🔎 Exploratory Data Analysis

Example visualization:

• Higher BMI is associated with higher medical charges.
• Smokers clearly have much higher charges compared to non-smokers.

---

📈 Regression Model

I built a multiple linear regression model including all variables (categorical variables converted to dummies).

Key results:

• R² = 0.75 → The model explains about 75% of the variance in medical charges.
• Statistically significant predictors:
  • age: Older individuals tend to have higher charges.
  • bmi: Higher BMI is linked to higher charges.
  • children: More children slightly increases charges.
  • smoker: Smoking has the largest effect, increasing charges by ~23,850 USD on average.
  • Some regional differences also appear significant.
• Not significant: Sex (male/female) does not have a meaningful effect.

---

🛠 Tools & Libraries

• Python (pandas, os, matplotlib, seaborn, statsmodels)
• Jupyter Notebook

---

📌 Key Learnings

• How to prepare categorical data using dummy variables.
• Building and interpreting linear regression models with statsmodels.
• Identifying which features are statistically significant in predicting outcomes.

---

🚀 Next Steps

• Try regularization techniques (Lasso, Ridge) to improve model robustness.
• Explore interaction effects (e.g., smoking × BMI).
• Deploy the notebook as an interactive dashboard (e.g., Streamlit).

---

This site is open source. Improve this page.