In this academic project, my teammate Karina Diana Templer and I developed an AI model to forecast house prices using supervised machine learning. To ground our work in a realistic scenario, we imagined a fictional real estate agency seeking to improve its pricing estimates. Using a rich dataset from Melbourne, Australia (sourced via Kaggle), our goal was to train and test predictive algorithms that could eventually be adapted for broader use across different markets. The model relies on historical housing data and features such as location, number of rooms, and local amenities—factors commonly used in real estate valuation.

Implementation approach:
We explored and compared two key supervised machine learning models: Multiple Linear Regression and Random Forest. The linear regression model allowed us to predict prices by analyzing the relationship between multiple property features and their selling prices. This model offered a clear, interpretable baseline. On the other hand, the Random Forest model, an ensemble method that combines multiple decision trees, achieved higher accuracy, with a stronger R-squared value and lower RMSE, making it a more robust choice for this dataset.

Beyond technical implementation, we also considered ethical implications such as bias in training data and the risk of perpetuating socioeconomic inequalities. Our design process included strategies to ensure transparency and fairness, especially as the model scales.

This model could be adapted for various real-world applications such as real estate investment analysis, property appraisal, development planning, and market research. With further training on diverse datasets, it has the potential to become a powerful tool for decision-making in the housing market.