The use of Machine Learning techniques to predict farm size change: an implementation in the Dutch Dairy sector

This paper investigates the use of several machine learning techniques in order to predict dairy farm size change in the Netherlands. The work presented is part of a larger effort to improve an agricultural model, called the Financial Economic Simulation (FES) model. The FES model simulates midterm financial economic development of farms, but until now it has not taken farm size change into account, which made it static, hence, sub-optimal when significant structural changes might occur in agriculture. In our work, we used data from the Dutch Farm Accountancy Data Network (FADN), covering the period between 2001 and 2009. After preprocessing the data, we built models using Multiple Linear Regression (MLR) and Neural Networks (NN), and measured model performance at various prediction periods (looking ahead one to eight years in time). Our results show that the chosen methods are able to predict farm size change effectively, and that prediction quality is best when the aim is to predict farm size four years ahead, for which we also provide a likely explanation.