Estimating residual value of heavy construction equipment using ensemble learning

Abstract

Knowing the right moment for the sale of used heavy construction equipment is important information for every construction company. The proposed methodology uses ensemble machine learning techniques to estimate the price (residual value) of used heavy equipment in both the present and the near future. Each machine in the model is represented with four groups of attributes: age and mechanical (describing the machine) and geographical and economic (describing the target market). The research suggests that the ensemble model based on random forest, light gradient boosting, and neural network members, as well as support vector regression as a decision unit, gives better estimates than the traditional regression or individual machine learning models. The model is built and verified on a large data set of 500,000 machines advertised in 50 US states from 1989 to 2012.