ZHPO-LightXBoost an integrated prediction model based on small samples for pesticide residues in crops

Abstract

Excessive dependence on and unreasonable use of pesticides in actual crop growth will lead to excessive pesticide residues and exacerbate environmental pollution. Therefore, the prediction of pesticide residues in crops is particularly important. In order to ensure the accuracy and generalization of the pesticide residue prediction model, a ZHPO-LightXBoost pesticide residue prediction model based on small sample datasets was proposed. The ZHPO algorithm was used to solve the problem of determining hyperparameters in the LightXBoost model. Comparative experiments conducted on four independently constructed datasets have shown that the proposed model achieves an increase in R² of 0.004–0.042, and achieves a reduction in MSE ranging from 0.0002 to 0.276 mg/kg, as well as a reduction in MAE ranging from 0.0005 to 0.181 mg/kg.