Machine Learning Framework for Dry Bean Classification in Smart Agriculture

Abstract

Looking at the exponentially growing population the effective and automatic farming techniques are important to provide major percentage of humanity's nutrition. To increase sustainable food production and reducing waste the optimized decision-making and classification of agricultural produce is crucial. To overcome the time-consuming manual inspection technique the automatic and cost-effective techniques of nutrient management, water utilization, and pest control are proposed. In this work, Naive Bayes, Multi-Layer Perceptron (MLP), and Kernel Support Vector Machine (K-SVM) models are applied to the dry beans data to multiclass classification of bean varieties. The efficacy of these algorithms is measured based on the standard evaluation parameters such as accuracy and F1_macro. The K-SVM classifier achieves the highest overall accuracy, with a value of 0.93 (93%), followed by MLP at 0.92 (92%) and Naive Bayes at 0.90(90%), as shown in the results section.