An end-to-end data analytics and machine learning project focused on predicting housing prices in King County, Washington (including Seattle). This project was developed as part of the IBM Data Analysis with Python professional certification.
The objective of this project is to analyze a comprehensive real estate dataset and build predictive models to estimate house market values based on key architectural and geographical features (such as square footage, number of bedrooms/bathrooms, view, grade, and location).
- Language: Python
- Data Manipulation: Pandas, NumPy
- Data Visualization: Matplotlib, Seaborn
- Machine Learning: Scikit-Learn
- Handled missing data points using statistical imputation (replacing missing values in key columns with the mean).
- Adjusted data structures and optimized variable data types for modeling.
- Conducted advanced correlation analysis to extract impactful features affecting house valuations.
- Developed categorical visualizations (e.g., box plots evaluating price distributions for waterfront vs. non-waterfront properties).
- Utilized statistical grouping (
groupby) to analyze regional price segments.
Developed, fine-tuned, and cross-validated multiple regression models to evaluate performance tradeoffs:
- Simple Linear Regression as a baseline model.
- Multiple Linear Regression using highly correlated features to capture multidimensional patterns.
- Polynomial Transformations (2nd order) to extract and learn non-linear relationships.
- Ridge Regression (L2 Regularization) optimized with hyperparameter tuning (
Alpha) to constraint coefficients, manage model complexity, and effectively mitigate overfitting. - Scikit-Learn Pipelines deployed to bundle preprocessing steps (such as data scaling/standardization) and model fitting into a unified, clean object.
- Strong correlations were identified between a house's living area square footage (
sqft_living), overall building grade, and its final market price. - Implementing regularization methods and advanced feature scaling significantly enhanced the model's capacity to generalize well on unseen test data.
Note: This project serves as a practical demonstration of regression analysis, feature engineering, and statistical modeling in real-world business scenarios.