Mahy et al. (2014) have developed a non-parametric methodology to predict land use choices of farmers in the context of the crop diversification measure. The methodology uses simulation at the micro level because crops cultivated by one farm cannot compensate for a lack of diversity of crops at another farm. A key difficulty of simulating at farm level is that the crop choice of an individual farmer is very difficult to predict because it depends on more factors than gross margins only, such as crop rotation, farmers experience, adaptability of machines and supply chain possibilities of the harvested products. The non-parametric methodology presented in this paper uses peer behaviour of farmers to identify choices of other farmers. This paper validates and compares the approach with improved versions on a regional case study in Flanders to show the possibilities and limitations of the methodology.