Real Estate Price Estimation with AI: A Hybrid Approach Combining Clustering and Machine Learning

Accurate price prediction in the real estate market is important for buyers, sellers, and investors. This study evaluates the performance of various machine learning models including AdaBoost, Gradient Boosting, k-Nearest Neighbors (kNN), Artificial Neural Networks, and Support Vector Machines (SVM) to predict house prices in Gaziantep, Turkey. Parameters such as number of rooms, square meters, building age, floor level, and neighborhood are taken as datasets from a real estate website. A hybrid study is conducted to improve the model performance by clustering analysis using the Simple K-Means algorithm in WEKA application and categorizing the data into groups according to the parameters. The clustered data served as input for Orange 3. Model performance is evaluated using metrics such as Mean Square Error (MSE), Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and R². The results show that AdaBoost consistently achieves the highest accuracy and reliability, followed by Gradient Boosting, which demonstrates strong generalization capabilities. While kNN provided moderate performance, Neural Networks and SVM performed poorly, showing high error measures and poor adaptability.