Investigations on Feature Ranking Algorithms for Date Fruit Genotype Detection

Post-harvest industrial processes serve as essential for preserving the economic worth of an agricultural crop. For the separation and categorization of date fruits to be more effective, novel techniques and innovations are required. This work aims to construct artificial intelligence (AI)-based classification algorithms that differ from existing separation strategies and are capable of delivering the necessary categorization. In this work, proposed framework developed with a system with many levels which involve extraction of features, selection of features, and classification. This study investigates the efficacy of several Support Vector Machine (SVM) and Neural Network (NN) classifiers. Further, to investigate the effectiveness of the feature ranking algorithms, Minimum Redundancy Maximum Relevance (MRMR), chi-square (Chi2), ReliefF and Kruskl_Wallis algorithms are used in this work. The performance is measured with SVM and NN classifiers at different set of features ranging from 34 to 20. The experimental results depicted that MRMR and Kruskl_Wallis ranking algorithms outperformed Chi2 and ReliefF algorithms.