From Housing Data to an Interactive ML App, Predicting Real Estate Prices with Python, Scikit-learn and Streamlit
Published by John Oruebi Eniekpo in Python & Machine Learning
Machine learning becomes more useful when a model can move beyond a notebook and become something people can interact with.
For this project, I developed an end-to-end real estate price prediction application using Python, Scikit-learn, Pandas, Streamlit and Plotly.
The objective was to build a regression model capable of estimating median house values from housing, demographic and geographical characteristics, while also creating a professional user interface where users can explore the underlying dataset, evaluate different machine-learning models and generate predictions interactively.
I used the California housing dataset, which contains more than 20,000 observations and includes variables such as geographical coordinates, housing age, number of rooms, population, household count, median income and proximity to the ocean.
Rather than treating the project only as a modelling exercise, I designed it as a complete machine-learning workflow covering:
1. Data acquisition and exploration 2. Data preprocessing 3. Missing-value treatment 4. Categorical-variable encoding 5. Train-test splitting 6. Regression modelling 7. Model evaluation 8. Pipeline development 9. Interactive prediction 10. Streamlit application development and visualization
The final application allows users to choose between Linear Regression, Random Forest Regression and Gradient Boosting Regression, review model-performance metrics and enter their own housing characteristics to generate an estimated median house value.