Floor heating customer prediction model based on random forest

Abstract

Nowadays floor heating service is increasingly attracting both residents in cold areas and gas companies for market profits. With the aggravation of market-oriented competition, the gas companies are actively seeking service transformation. It is of great significance to gas companies to be able to forecast those customers willing to use floor heating. In this paper, we establish a floor heating customer prediction model that helps indicate the potential customers using floor heating, based on analyzing existing floor heating customers’ behavior. The prediction model uses random forest. We exploit data coming from the actual running of a Shanghai based gas company. Experiments show that the random forest model has better performance than those using k-nearest neighbor (KNN) or logistic regression.