Predicting MEDV and NOX – Boston Housing Dataset

This project explores environmental and socioeconomic factors that influence two key variables in the Boston housing dataset:

• MEDV — Median value of owner-occupied homes
• NOX — Nitric oxide concentration (parts per 10 million)

Using correlation analysis and linear regression modeling, the notebooks identify the most influential features and evaluate model performance.

Contents

• Predicting_MEDV.ipynb: Regression analysis is used to identify the best predictors of MEDV, including RM (average number of rooms) and LSTAT (lower-status population).
• Predicting_NOX.ipynb: Regression analysis is used to identify the best predictors of NOX, focusing on AGE (older buildings) and DIS (distance to employment centers).
• PDF exports are included for each notebook

Tools & Libraries

• pandas, numpy
• seaborn, matplotlib
• statsmodels (OLS regression)

Key Insights

• LSTAT and RM show the strongest relationships with MEDV
• AGE and DIS have significant predictive power for NOX
• All models are statistically significant with strong R² values

Data Source

Boston Housing Dataset (originally from UCI, accessible via data/boston.csv)

License

MIT License — see the LICENSE file for details.