Performance Analysis of Feature Selection Algorithms in the Classification of Dry Beans using KNN and Neural Networks

Post-harvest industrial operations are crucial to the preservation of the economic value of an agriculture product. For dry bean separation and classification to be more efficient, new methods and technologies are necessary. Different dry beans appeal to various markets, which increases the need for categorising beans. The purpose of this study is to develop machine learning (ML)-based classification models that vary from existing separation approaches and are capable of providing the required categorization. In this study, a multi-level system that includes feature extraction, feature selection, and classification were designed. This research analyses the performance of a range of k-nearest neighbours (KNN) and neural network (NN) classifiers. From a dataset including 13,611 samples, a set of 16 features is extracted. In addition, the minimal redundancy maximum relevance (MRMR) and ReliefF feature ranking algorithms are employed to choose the optimal set of features for training the ML models. The results of the experiments indicated that the proposed technique had a classification efficiency of 93.9%. The study demonstrates that the suggested models are superior to existing techniques.