Scholarlink

<p>Effective quality assessment methods are critical for ensuring the safety of beef, a globally significant protein source. This study introduces a rapid, non-destructive approach leveraging the gcForest algorithm to evaluate beef freshness through bioimpedance spectroscopy. Fresh beef samples (hind leg, collected ≤ 48&#xa0;h post-slaughter) were stored at 0&#xa0;°C, 4&#xa0;°C, 10&#xa0;°C, and 25&#xa0;°C to simulate real-world conditions. Total Volatile Basic Nitrogen (TVB-N) levels, a key freshness indicator, were correlated with impedance measurements acquired via two- and four-electrode configurations across 10 excitation frequencies (100&#xa0;Hz–150&#xa0;kHz). To optimize model robustness, the framework integrated OCSVM-based outlier detection, LASSO regression for feature selection, and PCA for dimensionality reduction. The gcForest algorithm, employing a deep forest architecture with multi-granularity scanning, demonstrated superior performance over conventional machine learning models (Decision Tree, Random Forest, AdaBoost, etc.). Notably, the four-electrode configuration achieved exceptional accuracy, with evaluation metrics (MAE: 0.8150, MSE: 0.9611, RMSE: 0.9803, R²: 0.9208) outperforming the two-electrode setup (MAE: 0.8281, RMSE: 0.9902, R²: 0.8953), attributable to enhanced signal fidelity in capturing microbial-driven impedance changes.This non-invasive method enables precise, real-time freshness assessment, providing actionable insights for food safety systems. Its operational simplicity and reliability position it as a transformative tool for meat quality control, with direct applications in supply chain monitoring and consumer health protection.</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

An impedance-based beef freshness detection method using GcForest learning model

  • 期刊论文

  • 发表日期：

    2025年4月9日

查看全文
  • Xinxing Li,
  • Shu Li,
  • Jianjun Tian,
  • Buwen Liang

摘要

Effective quality assessment methods are critical for ensuring the safety of beef, a globally significant protein source. This study introduces a rapid, non-destructive approach leveraging the gcForest algorithm to evaluate beef freshness through bioimpedance spectroscopy. Fresh beef samples (hind leg, collected ≤ 48 h post-slaughter) were stored at 0 °C, 4 °C, 10 °C, and 25 °C to simulate real-world conditions. Total Volatile Basic Nitrogen (TVB-N) levels, a key freshness indicator, were correlated with impedance measurements acquired via two- and four-electrode configurations across 10 excitation frequencies (100 Hz–150 kHz). To optimize model robustness, the framework integrated OCSVM-based outlier detection, LASSO regression for feature selection, and PCA for dimensionality reduction. The gcForest algorithm, employing a deep forest architecture with multi-granularity scanning, demonstrated superior performance over conventional machine learning models (Decision Tree, Random Forest, AdaBoost, etc.). Notably, the four-electrode configuration achieved exceptional accuracy, with evaluation metrics (MAE: 0.8150, MSE: 0.9611, RMSE: 0.9803, R²: 0.9208) outperforming the two-electrode setup (MAE: 0.8281, RMSE: 0.9902, R²: 0.8953), attributable to enhanced signal fidelity in capturing microbial-driven impedance changes.This non-invasive method enables precise, real-time freshness assessment, providing actionable insights for food safety systems. Its operational simplicity and reliability position it as a transformative tool for meat quality control, with direct applications in supply chain monitoring and consumer health protection.