Enhancing Fruit Maturity Detection using Convolutional Neural Networks Algorithm Compared with Naive Bayes Algorithm

Abstract

This study aims to compare the accuracy of the fruit maturity detection enhancement using Convolutional Neural Networks (CNNs) and Naive Bayes Algorithm, with a specific focus on various methods. This research also evaluates their effectiveness in Enhancing Fruit Maturity Detection. Using G*Power parameters of 0.8 for each group, 0.07 for alpha, and 0.2 for beta, the total sample size is calculated as 10,000 (5,000 samples in group 1 and 5,000 in group 2). To improve results, synthetic datasets were created. The Convolutional Neural Networks was implemented, and configured with Naive Bayes in deep learning. The selection of the most suitable approach is based on the outcomes derived from the SPSS statistical analysis. After evaluating both algorithms, it became evident that CNN outperformed Naïve Bayes, exhibiting a performance accuracy of 81.56% versus 54.79%. The sample T-test indicated no significant difference between CNN and Naïve Bayes, with a p-value of 0.048 (p < 0.05). This suggests that Convolutional Neural Networks can handle datasets of varying sizes effectively, while Naïve Bayes performs reasonably well with smaller datasets and can be trained quickly.