Real estate analytics and price prediction using automated machine learning

Navigating the Malaysian real estate market is increasingly complex due to the dynamic interplay of localised socioeconomic factors and high-dimensional variables. Addressing the demand for objective decision-support tools, this research develops a comprehensive analytics and price prediction framework tailored for four primary economic hubs: Selangor, Kuala Lumpur, Penang, and Johor. Diverging from conventional manual modelling, the methodology employs an Automated Machine Learning (Auto ML) approach via the PyCaret library to systematically explore feature spaces while minimising human bias during algorithm selection. Evaluation of various regression architectures utilised standard benchmarks, specifically Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and the Coefficient of Determination R2. Empirical results indicate that ensemble-based models consistently exceed the performance of alternative approaches; Random Forest achieved the highest accuracy for Selangor, Penang, and Johor, while the Extra Trees Regressor emerged as the superior model for Kuala Lumpur. Furthermore, a unified multi-regional model demonstrated exceptional generalisation capabilities, recording an R2 of 0.893, an RMSE of RM68,497.17, and an MAE of RM46,104.28 on a combined dataset. The primary contribution of this work lies in the seamless architectural integration of these high-performance models into a real-time, web-based dashboard, transitioning the research from static academic analysis to a dynamic, user-facing system. Ultimately, these findings confirm that an Auto ML-driven framework simplifies complex analytical workflows and delivers the precision required to support data-driven decision-making and transparent property valuation in a developing economy.