The sales rate is very important information for condominium design and development processes. However, predicting it accurately requires substantial expertise and experience. This research investigated the use of a Deep Learning model to predict condominium sales rates and explored various determinants that influence sales rates. The research was done by (1) identifying the determinant factors from the literature review, (2) collecting 199 data from market survey reports, (3) creating a Deep Learning model that can predict the sales rate of a condominium by its determinant factors, and (4) verifying the model by checking its coefficient of determination (R2 value). The research results revealed the three-layered network (34 input nodes, 19 hidden layer nodes and one output node) with R2 of 0.703 and RMSE of 0.539, showing the model has enough accuracy for planning purposes. The determinant factors comprise both internal and external factors, which can be divided into six groups: (1) Price-related factors, (2) Entrepreneur-related factors, (3) Room-related factors, (4) Project structure-related factors, (5) Common-area-related factors, and (6) Location-related factors. By understanding the influence of this set of data on project sales rate, the project stakeholders can efficiently contribute to the project’s success. In addition, the results showed the high accuracy of the Deep Learning model even in the case of a limited amount of data.
Read full abstract