Modeling of farmers’ production performance portrait based on GASA-WFCM clustering

Abstract

To identify the typical characteristics of farmers with different production performance, this paper constructs a farmer portrait model from the perspective of production performance segmentation. Firstly, we constructed the evaluation index system of production performance and quantified grape farmers’ production performance. Then, an optimized clustering algorithm (GASA-WFCM) was proposed to cluster the grape farmers based on their production performance, which was based on fuzzy C-mean algorithm, and incorporating the genetic simulated annealing algorithm to overcome the problem of sensitive initial clustering centers and tendency to fall into local optimum, and the ReliefF algorithm was also integrated to strengthen the contribution ability of effective features to clustering. Finally, we designed a labeling system and feature transformation quantification method, and combined with the visualization effects to realize the description of grape farmers’ portraits. The results of models training indicated that the convergence speed and division accuracy of the optimized clustering algorithm were significantly improved. By comparing the clustering results on the typical UCI datasets, it was demonstrated that GASA-WFCM algorithm improved the accuracy, NMI and ARI by 3.0%, 7.2% and 6.1%, respectively. Moreover, the segmentation method was applied in grape farmers dataset, and the findings revealed that Chinese fresh grape farmers may be clustered into three groups, namely high-yield-driven grape farmers, high-input-driven grape farmers and high -income grape farmers, respectively. This paper provided a set of methods of users clustering analysis and portrait from a theoretical point of view. The findings and suggestions can help provide accurate and focused decision support for improving the effect of agricultural technology promotion and farmers’ production performance.