A Decision Support Tool for Buying Farm Tractors, Based on Predictive Analytics

Abstract

Data science can help farmers when making a decision about tractor purchase. Buying a tractor represents a big investment for farmers, and price is one of their main concerns. This study presents the development of a web-based decision support tool (DST) that calculate the price of new and second-hand tractors, for the purpose of providing the decision-maker some information that will lead him to the final decision. The tool makes use of different algorithms based on predictive analytics methodologies. The dataset has information about 227 different observations of new tractors and 1003 of second-hand tractors, from different European countries. During the study, the prices of new and used tractor were modeled, testing parametric and non-parametric regression models with different segmentations and predictor variables. Non parametric models includes regression trees, support vector machines, ensembles of regression trees, Gaussian process, and neural networks. In both cases, for predicting the prices of new and second-hand tractors, adjusted R2 higher than 0.99 were achieved. The models developed were implemented in the DST which is fully operative, available in Internet, and free to use.