Accurate fruit classification, an important feature of agricultural production, has motivated the development of advanced machine-learning algorithms. Traditional classification methods often struggle to capture complex relationships in such data sets. This paper presents a fuzzy kernel function for a multi-layer extreme learning machine (MLELM), which is explicitly designed to classify random fruit datasets. The fuzzy kernel-based MLELM model is developed by integrating fuzzy set theory with multi-layer kernel extreme learning machine models. The method incorporates feature information extracted using fuzzy set theory into a multi-layer extreme learning machine where its kernel function is designed using its fuzzy lower and upper approximations of fresh and rotten classes. The fuzzy logic handles uncertain data and provides adaptability to complex datasets. The performance of the model is compared with support vector machine (SVM) and decision tree which exhibit higher accuracy in fruit classification. Experimental results show that the proposed model achieves 68.48% testing accuracy, outperforming SVM, decision tree, ELM, and MLELM, which are 64.09% and 49%, 66.12%, and 67.25% respectively. This study contributes to the field by providing a new framework in fruit classification, using fuzzy set theory and kernel function integration. The proposed model shows promise as a reliable tool to evaluate the quality of the fruit, which enables improvement in cultivation strategies.