Prediction of Selenium-Enriched Crop Zones in Xiaoyan Town Using Fuzzy Logic and Machine Learning Approaches

Abstract

Selenium-rich foods play a crucial role in human health and hold significant economic value for agricultural products. However, many regions in China are experiencing selenium deficiency, which has led to an increased demand for Se-rich agricultural products. This study focused on Nanzhang County, a key area within the “Organic Valley” of Hubei Province, China. We employed fuzzy weights-of-evidence, backpropagation neural network, and support vector regression models to predict optimal planting zones for Selenium-rich crops. A comparative analysis indicated that the backpropagation neural network model provided the highest prediction accuracy (R2 = 0.77), identifying Selenium-rich crop zones covering 68.87% of the aera, where Selenium-rich crops made up 86.67% of all samples. Notably, the backpropagation neural network yielded excellent performance for rice and rapeseed, with R2 values of 0.95 and 0.99, respectively. The findings also indicate that the Selenium content in crops is closely linked to Selenium levels in the soil and is significantly influenced by synergistic and antagonistic interactions with other elements. This study provides scientific support for the cultivation of selenium-rich crops. It plays a positive role in promoting the development of the local selenium-rich industry and the sustainable utilization of soil selenium resources.