Prediction of Excessive Cadmium in Rice Based on Weighted Bayesian Fusion Model

Soil Cd pollution directly affects the safety of agricultural products. Frequent Cd pollution incidents have repeatedly reminded that soil Cd pollution in China is not optimistic. In this paper, the linear regression prediction based on SOM, pH, and soil Cd in the data was firstly carried out. The fitting degree R2 was 0.109. The model was not accurate enough. Therefore, this paper converts the study of rice Cd beyond the predicted regression problem into a classification problem. According to the Chinese national standard GB2762-2012 food contaminant limit, the content of Cd in rice should not exceed 0.2 mg/kg. Finally, the adaptive classification algorithm based on the weighted Bayesian fusion model is compared with the classical classifier model (SVM, RF). The adaptive classifier is better than the classical classifier in F1 and accuracy index. The adaptive classifier model obtained from the above studies can not only guide the classification of rice Cd but also save huge costs in the monitoring and prevention of Cd pollution.