Cost-sensitive deep forest for price prediction

Abstract

For many real-world applications, predicting a price range is more practical and desirable than predicting a concrete value. In this case, price prediction can be regarded as a classification problem. Although deep forest is recognized as the best solution to many classification problems, a crucial issue limits its direct application to price prediction, i.e., it treated all the misclassifications equally no matter how far away they are from the real classes, since their impacts on the accuracy are the same. This is unreasonable to price prediction as the misclassification should be as close to the real price range as possible even if they have to be wrongly classified. To address this issue, we propose a cost-sensitive deep forest for price prediction, which maintains the high accuracy of deep forest, and propels the misclassifications to be closer to the real price range to reduce the cost of misclassifications. To make the classification more meaningful, we develop a discretization method to pre-define the classes of price, by modifying the conventional K-means method. The experimental results based on multiple real-world datasets (i.e., car sharing, house renting and real estate selling) show that, the cost-sensitive deep forest can significantly reduce the cost in comparison with the conventional deep forest and other baselines, while keeping satisfactory accuracy.