COMPARATIVE EVALUATION OF SVM AND RANDOM FOREST FOR ORANGE FRUIT DISEASE CLASSIFICATION

Industrialists in food production are constantly looking for methods of improving efficiency in their workforce. It is commonly known that as time has progressed, industrialists have keenly explored the employment of technology and AI into their workforce in anticipation of decreasing labor costs and increasing productivity. A prevalent area where food production can improve in terms of efficiency is food health detection. During mass production of food like oranges, it takes a lot of time for human labor to inspect and approve the health of oranges. Consequently, it would be beneficial if AI could filter this process through image classification to preliminarily check the health of oranges such that it makes the production process faster. Subsequently, this paper explores the performance of support vector machines compared to random forests in orange disease prediction. In the methodology, a dataset from Kaggle was used that comprised of images of oranges that were healthy, those with melanoses, and those with citrus canker. This way the two models were compared and evaluated. At the end, the support vector machine had better scores at 90% compared to random forests.