Digital Image-based Classification of Clove Quality using Naïve Bayes Algorithm

Abstract

At present, clove cultivation is increasingly in demand, especially by farmers because it is easy to maintain and the selling price is high. Researchers conducted observations in Tana Toa Village on how the processing of cloves after harvesting is drying them in the sun until they turn brown and shrink. After that, farmers select dried cloves and distinguish between good and bad quality. One way for farmers and traders to determine the quality of cloves is by visually inspecting the size and color. One of the disadvantages of this manual classification process is that each person can look at the same material in bulk in different ways depending on the situation or individual weak points. The aim of this research is to help farmers produce high quality cloves that will ultimately produce favorable results on their economy. With digital image-based methods and Naive Bayes, this process can be done quickly and efficiently, reducing operational costs and labor time. The Naive Bayes algorithm is able to process data more thoroughly than humans, especially if the image quality and features used for classification are optimized. This reduces human errors that may occur during manual processing. The results of this study are, Gaussian Naive Bayes testing has an accuracy of 0.82. Bernoulli naïve bayes has an accuracy of 0.69, Complement naïve bayes and multinomial naïve bayes each have an accuracy of 0.89. This shows that they affect the accuracy rate of clove quality effectively