<![CDATA[This research develops a house price prediction system in Surabaya using the Backpropagation Neural Network (BPNN) method. The dataset was obtained through web scraping of property listings, resulting in 3,435 records with 52 attributes. To improve stability, the target variable (house price) was transformed using natural logarithms. Several neural network architectures were tested, and the best configuration [32, 64, 32] achieved Mean Absolute Error (MAE) of 0.3125, Root Mean Squared Error (RMSE) of 0.4201, R² of 0.7138, and Mean Absolute Percentage Error (MAPE) of 1.46%. A multi-run evaluation of 20 iterations confirmed consistency of results. The model was implemented as a web-based application using Flask, allowing users to predict house prices in real-time. This research shows that BPNN is reliable for property price forecasting and can support decision-making in the housing market.]]>
Copyrights © 2025
