Scholarlink

Aims <p>Accurate prediction of bioaccumulation and risk of cadmium (Cd) in wheat is important for assessing the safe utilization and risk management of Cd-contaminated soils.</p> Methods <p>This study combined machine learning (ML) with Bayesian risk prediction models to quantify key factors and predict Cd risk for two main wheat varieties on a county in southwest China.</p> Results <p>Three ML models (RF, XGB, and GBDT) were used to predict wheat Cd (wCd) contents using paired soil-wheat samples (<i>n</i> = 96). Additionally, feature importance analysis to identify the key factors on wCd contents based on 11 factors in 3 categories. Bayesian risk prediction model was used to predict the risk of wCd exceeding the food safety standard (0.1&#xa0;mg&#xa0;kg<sup>−1</sup>). The feature importance analysis revealed that the top three influencing factors for both wheat varieties were aZn (37.72% and 56.74%, respectively), tCd (27.37% and 13.53%), and aCd (27.25% and 11.49%).Using tCd and aCd as variables for risk prediction, the results showed that when the contents of tCd and aCd were 0.66, 0.72, 0.77&#xa0;mg&#xa0;kg<sup>−1</sup> and 0.41, 0.48, 0.53&#xa0;mg&#xa0;kg<sup>−1</sup>, respectively, the wCd risk reaches 10%, 50%, and 90%, respectively.</p> Conclusions <p>Our study provides valuable insights for predicting Cd levels and probability of exceeding risk in wheat, thus helping to ensure food safety.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A novel quantitative approach for factor identification and risk prediction of cadmium accumulation in wheat using machine learning and Bayesian models

  • 期刊论文

  • 发表日期：

    2026年1月15日

查看全文
  • Yakun Wang,
  • Zhuo Zhang,
  • Chouyuan Liang,
  • Haochong Huang,
  • Kening Wu,
  • Huafu Zhao,
  • Yuxian Shangguan

摘要

Aims

Accurate prediction of bioaccumulation and risk of cadmium (Cd) in wheat is important for assessing the safe utilization and risk management of Cd-contaminated soils.

Methods

This study combined machine learning (ML) with Bayesian risk prediction models to quantify key factors and predict Cd risk for two main wheat varieties on a county in southwest China.

Results

Three ML models (RF, XGB, and GBDT) were used to predict wheat Cd (wCd) contents using paired soil-wheat samples (n = 96). Additionally, feature importance analysis to identify the key factors on wCd contents based on 11 factors in 3 categories. Bayesian risk prediction model was used to predict the risk of wCd exceeding the food safety standard (0.1 mg kg−1). The feature importance analysis revealed that the top three influencing factors for both wheat varieties were aZn (37.72% and 56.74%, respectively), tCd (27.37% and 13.53%), and aCd (27.25% and 11.49%).Using tCd and aCd as variables for risk prediction, the results showed that when the contents of tCd and aCd were 0.66, 0.72, 0.77 mg kg−1 and 0.41, 0.48, 0.53 mg kg−1, respectively, the wCd risk reaches 10%, 50%, and 90%, respectively.

Conclusions

Our study provides valuable insights for predicting Cd levels and probability of exceeding risk in wheat, thus helping to ensure food safety.