Explainable Machine Learning Framework for Milk Quality Grading

Bekir Çetintav; Ahmet Yalçın

Article Details

Explainable Machine Learning Framework for Milk Quality Grading

Authors:

Bekir Çetintav | Ahmet Yalçın

Publisher:

Afyon Kocatepe University

Main Topics:

Veteriner Gıda Hijyeni ve Teknolojisi | Zootekni | Genetik ve Biyoistatistik

Journal:

Kocatepe Veteriner Dergisi Vol.18 Issue 3, pp.227-235

Document Type:

Article

Publication Date:

3/2025

Document Language:

English

Document ID (Org No.):

2001354168

pISSN:

1308-1594

Keywords:

Abstract:

 This study introduces an explainable machine learning framework for milk quality grading, combining high predictive performance with transparency and practicality. Utilizing Random Forest and HistGradientBoost models, alongside interpretability techniques like Permutation Feature Importance and LIME, the framework achieves robust classification while providing actionable insights. Global explanations identify pH and Temperature as critical factors, highlighting their significance in real-time monitoring and microbial control. Local explanations, based on the two presented examples, demonstrate the practical utility of individual predictions, offering targeted interventions such as optimizing storage conditions or addressing contamination risks. By bridging the gap between predictive accuracy and interpretability, this framework not only enhances trust and usability for stakeholders but also establishes a new perspective for integrating AI-driven quality control systems into the dairy industry.

 This study introduces an explainable machine learning framework for milk quality grading, combining high predictive performance with transparency and practicality. Utilizing Random Forest and HistGradientBoost models, alongside interpretability techniques like Permutation Feature Importance and LIME, the framework achieves robust classification while providing actionable insights. Global explanations identify pH and Temperature as critical factors, highlighting their significance in real-time monitoring and microbial control. Local explanations, based on the two presented examples, demonstrate the practical utility of individual predictions, offering targeted interventions such as optimizing storage conditions or addressing contamination risks. By bridging the gap between predictive accuracy and interpretability, this framework not only enhances trust and usability for stakeholders but also establishes a new perspective for integrating AI-driven quality control systems into the dairy industry.

 This study introduces an explainable machine learning framework for milk quality grading, combining high predictive performance with transparency and practicality. Utilizing Random Forest and HistGradientBoost models, alongside interpretability techniques like Permutation Feature Importance and LIME, the framework achieves robust classification while providing actionable insights. Global explanations identify pH and Temperature as critical factors, highlighting their significance in real-time monitoring and microbial control. Local explanations, based on the two presented examples, demonstrate the practical utility of individual predictions, offering targeted interventions such as optimizing storage conditions or addressing contamination risks. By bridging the gap between predictive accuracy and interpretability, this framework not only enhances trust and usability for stakeholders but also establishes a new perspective for integrating AI-driven quality control systems into the dairy industry.

 Bu çalışma, süt kalitesinin değerlendirilmesinde yüksek tahmin doğruluğunu şeffaflık ve kullanışlılık ile birleştiren açıklanabilir bir makine öğrenimi yaklaşımı sunmaktadır. Random Forest ve HistGradientBoost modellerinin yanı sıra Permutasyon Feature Importanee ve LIME gibi yorumlanabilirlik tekniklerini kullanan bu yaklaşım, güçlü bir sınıflandırma performansı sağlarken uygulanabilir içgörüler de sunmaktadır. Global yorumlanabilirlik sonuçları, pH ve Sıcaklık gibi kritik faktörleri belirleyerek gerçek zamanlı izleme ve mikrobiyal kontroldeki önemlerini vurgulamaktadır. Yerel yorumlanabilirlik sonuçları ise, sunulan 2 örnek üzerinden, bireysel tahminlerin pratik faydasını göstererek depolama koşullarının optimize edilmesi veya kontaminasyon risklerinin ele alınması gibi hedefe yönelik müdahalelere olanak tanımaktadır. Tahmin doğruluğu ile yorumlanabilirlik arasındaki boşluğu kapatan bu yaklaşım, yalnızca paydaşlar için güven ve kullanılabilirliği artırmakla kalmayıp, aynı zamanda AI destekli kalite kontrol sistemlerinin süt endüstrisine entegrasyonu için yeni bir perspektif sunmaktadır.

Article Details

Explainable Machine Learning Framework for Milk Quality Grading

Authors:

Bekir Çetintav | Ahmet Yalçın

Publisher:

Afyon Kocatepe University

Main Topics:

Veteriner Gıda Hijyeni ve Teknolojisi | Zootekni | Genetik ve Biyoistatistik

Journal:

Kocatepe Veteriner Dergisi Vol.18 Issue 3, pp.227-235

Document Type:

Article

Publication Date:

3/2025

Document Language:

English

Document ID (Org No.):

2001354168

pISSN:

1308-1594

Milk quality

Machine learning

Explainable AI (XAI)

Dairy Industry

Veterinary Food SafetySüt kalitesi

Makine öğrenimi

Açıklanabilir Yapay Zeka (XAI)

Süt endüstrisi

Veteriner gıda güvenliği

Issues

Article Details

Authors: Bekir Çetintav | Ahmet Yalçın

Publisher: Afyon Kocatepe University

Main Topics: Veteriner Gıda Hijyeni ve Teknolojisi | Zootekni | Genetik ve Biyoistatistik

Journal: Kocatepe Veteriner Dergisi Vol.18 Issue 3, pp.227-235

Document Type: Article

Publication Date: 3/2025

Document Language: English

Document ID (Org No.): 2001354168

pISSN: 1308-1594

Issues

Authors:

Bekir Çetintav | Ahmet Yalçın

Publisher:

Afyon Kocatepe University

Main Topics:

Veteriner Gıda Hijyeni ve Teknolojisi | Zootekni | Genetik ve Biyoistatistik

Journal:

Kocatepe Veteriner Dergisi Vol.18 Issue 3, pp.227-235

Document Type:

Article

Publication Date:

3/2025

Document Language:

English

Document ID (Org No.):

2001354168

pISSN:

1308-1594