Exploring the Relationship Between Rent and Flat Prices Through Random Forest and Grid Search

This paper explores the critical relationship between flat rents and prices, an important metric reflecting the state of the housing market and its underlying economic factors. The study introduces predictive models based on Random Forest algorithms, refined through Grid Search optimisation, to provide accurate market forecasts. Such predictive capabilities are crucial for policymakers, investors, and developers, who require reliable data for strategic decision-making and policy formulation. The research methodology employs Random Forest for its robust handling of complex datasets, with Grid Search utilised to calibrate the model parameters for enhanced precision. The resulting models are rigorously evaluated, achieving a training accuracy of 0.948 and testing accuracies of 0.745 for price prediction and 0.885 for rent prediction, indicating predictive solid performance. Low Log Loss metrics further validate the models’ reliability. The study identifies the model’s strengths and areas for refinement through detailed visualisations—accuracy comparisons, correlation heatmaps, and residual distribution plots. The residual plots, in particular, indicate a better generalisation capability in rent forecasting compared to flat price predictions, which exhibit minor overfitting. The findings provide actionable insights, demonstrating the efficacy of Random Forest and Grid Search in capturing the complex dynamics of flat rents and prices. The success of these models indicates their potential for broader application across different regions and market conditions, suggesting a valuable asset for stakeholders in achieving a balanced and fair housing market. This paper aims to contribute to the literature by offering a novel, data-driven perspective on market analysis and economic forecasting within the real estate sector