Real Estate Price Prediction Using Machine Learning

One of the significant factors that influence house valuation is its exterior and interior design. Realtors gather house images, arrange and label the pictures with their respective categories (bathroom, living room, bedroom, etc.) before eventually engaging with the clients. This manual annotation involves complex and very daunting tasks for a large volume of images, which may become unrealistic. In this study, a framework is developed using deep convolutional neural networks with different ResNet families (ResNet-50, ResNet-100 and Wide-ResNet) backbone to automatically classify different classes on real estate images and a shallow convolutional neural network for the price prediction. The performance of the models was evaluated on a real estate image dataset collected from Northern Cyprus. The developed system in this research can serve as an essential search tool in finding the right home or rental.