Accurate classification of opium poppy ( Papaver somniferum L.) genotypes based on phenotypic variation is essential for efficient breeding programs targeting specific traits such as flower and seed color. Opium poppy populations exhibit considerable variation in their morpho-phenological traits, particularly in flower and seed color, which is the focus of the present study. Descriptive plant measurements used to classify flower and seed color included germination time, flowering time, maturity time, plant height, number of capsules per plant, capsule length, capsule width, stigma rays, seed/capsule percentage, capsule yield per plant, seed yield per plant, capsule yield, seed yield, 1000-seed weight, morphine content, noscapine content, morphine yield, and noscapine yield. The opium poppy dataset ( nrow = 200) was divided into a training subset (70%) ( nrow = 140) and a test subset (30%) ( nrow = 60). ML algorithms, including Naïve Bayes, Support Vector Machine (SVM), weighted k-nearest neighbors (k-NN), Learning Vector Quantization (LVQ), and two decision-tree-based classification algorithms, bagging CART and Random Forest, were used for classification. According to the classification of flower and seed colors, the classification accuracies of Naïve Bayes, SVM, Weighted k-NN, and LVQ were found to be 95.00%, 91.67%, 78.33%, and 60.00% for flower color, and 78.33%, 91.67%, 50.00%, and 50.00%, respectively. Naïve Bayes and SVM classification yielded more accurate results for FC and SC, respectively.
Özgen et al. (Mon,) studied this question.