Comparative accuracy of different machine learning classifiers for characterizing varieties of pulse crops

Important machine learning classifiers viz., Logistic Regression, Linear Discriminant Analysis, K-Nearest Neighbor's and Naïve-Bayes were subjected for studing their accuracy, precision and recall accommodating 100 dataset each of 12 varieties (Awrodhi, C 235, JG 14, K 850, KWR 108, Pragati, PG 186, Pusa 362, Radhey, Shubhra, Udai, Ujjawal) of chickpea, 10 of lentil (DPL 62, HUL 57, IPL 316, K 75, NDL 1, PL 6, PL 8, PL 406, Pusa Vaibhav, Shekhar), 11 of fieldpea (Adarsh, Aman, Arkel, Azad pea 1, HUDP 15, Indra, IPF 4–9, IPFD 10–12, Prakash, Rachna, Vikash) 11 moong (PDM 139, HUM 12, HUM 16, IPM 02-3, IPM 02-4, Meha, NDM 1, Pant Moong 6, Pusa Vishal, Samrat, Sweta) and 11 of urdbean (Azad 1, Azad 2, Azad 3, IPU 02–43, NDU 1, PU 31, PU 40, Shekhar 1, Shekhar 2, T 9, Uttara) on the basis of their most important metric traits viz., plant height, size of leaf/leaflets, number of branches per plant, days to 50 per cent flowering, number of pods per plant, pod length, number of seed per pod and seed size in order to find out comparatively the best one model for characterizing the varieties. The average accuracy of Logistic Regression, Linear Discriminant Analysis, K-Nearest Neighbor's and Naïve-Bayes were varied over the pulse crops. The precision and recall of test data set of all crops varieties were 100%. The K-NN model was thus found to be out performed over other models under studied and could therefore effectively be utilized for characterizing, classifying and/or identifying the varieties of pulse crops.