Application of the Backpropagation ANN to Assess the Adoption Level of Farmers to Integrated Pest Management in the Province of Soc Trang (Vietnam)

Abstract

The integrated pest management (IPM) program was implemented in 2015 and 2016 in the province of Soc Trang. The research question is whether Artificial Neural Networks (ANNs) with pattern recognition can be useful for classifying farmers for a more realistic assessment of the performance of an IPM program. To evaluate the performance of the program, three datasets were collected, including dataset S1i with 450 farmers interviewed before conducting the IPM program, S2i with 250 farmers in the pilot area (communes/villages), and S3i with 50 farmers outside the pilot area. The conventional statistical assessment method (CAM) assumes that all farmers in each dataset behave similarly related to IPM concerning the seed, spray frequency, and dosage. This means that the original datasets were used to estimate the required statistical parameters. Thus, the traditional approach wastes information hidden in all surveyed data. Based on ANN, we can classify and determine the percentage of farmers in the six groups or the level of IPM adoption (3 neutral groups and 3 active groups) as well as the actual benefits of the IPM program. ANN-based assessment method (ANN-M) has been proven to be better than CAM in evaluating the performance of the project.