Feature-driven fruit classification using leveraging machine learning algorithms and nano semiconductor sensors

The proposed work compares machine learning algorithms for fruit classification using apples, mandarins, oranges, and lemons. The goal is to identify the most accurate and precision-score algorithm. To assemble a robust dataset, we purchased several dozen oranges, lemons, and apples of various subtypes and meticulously recorded their mass, width, height, and color score using nano semiconductor sensors. Using this recorded data statistical analysis is conducted for identifying the accurate fruit classification machine learning method. The accurate prediction rate is determined by subtracting the actual and anticipated values. K-Nearest Neighbors (KNN) shows superior performance, achieving accuracies of 0.989, 0981 and 0.979 on training, validation and testing sets. The performance of the KNN algorithm combined with the W-H-CS feature combination technique is highly dependent on the choice of k and relevance of the selected features.