The Classification of Germination Parameters Under Different Salinity Conditions Using Machine Learning Approaches
Türk Tarım - Gıda Bilim ve Teknoloji dergisi, cilt.14, sa.7, ss.1963-1970, 2026 (TRDizin)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 14 Sayı: 7
- Basım Tarihi: 2026
- Doi Numarası: 10.24925/turjaf.v14i7.1963-1970.8826
- Dergi Adı: Türk Tarım - Gıda Bilim ve Teknoloji dergisi
- Derginin Tarandığı İndeksler: CAB Abstracts, Central & Eastern European Academic Source (CEEAS), Directory of Open Access Journals, TR DİZİN (ULAKBİM)
- Sayfa Sayıları: ss.1963-1970
- Açık Arşiv Koleksiyonu: AVESİS Açık Erişim Koleksiyonu
- Hatay Mustafa Kemal Üniversitesi Adresli: Evet
Özet
Background: Soil salinity is a major abiotic stress affecting maize (Zea mays L.) production, leading to reduced germination and yield. Traditional breeding methods are resource-intensive and limited by environmental variability. This study explores machine learning (ML) algorithms to classify salinity stress levels based on germination parameters, aiming to enhance breeding efficiency for salt-tolerant genotypes. Methods: Germination data from 39 maize genotypes were collected under three salinity conditions (0 mM, 100 mM, 200 mM NaCl). Parameters included T10, T50, T90, T10-90, MGT, and Gmax. Data preprocessing involved normalization and feature selection using Correlation-based Feature Selection (CFS). Five ML models BayesNet, Decision Table, Bagging, J48, and Random Forest were evaluated using 10-fold cross-validation, with performance metrics of accuracy, precision, and F-measure. Results: Using all features, Random Forest achieved the highest accuracy (92.02%), followed by BayesNet (89.74%). After CFS (removing genotype and T10-90), BayesNet performed best (89.74%), while Random Forest slightly declined (87.46%). All models exceeded 87% accuracy, demonstrating robust classification of salinity levels. Conclusions: ML models effectively classify salinity stress in maize germination, even without certain parameters, supporting rapid genotype selection. This approach offers a scalable tool for improving agricultural resilience against salinization.