KLASIFIKASI JAMUR KONSUMSI MENGGUNAKAN RANDOM FOREST DENGAN FITUR HUE, SATURATION, VALUE (HSV) DAN HISTOGRAM OF ORIENTED GRADIENT (HOG)

Abstract

Mushrooms have a variety of shapes, sizes, and colors that make them important in the food and health sectors. Five types of edible mushrooms are commonly found, namely button mushrooms, straw mushrooms, shimeji mushrooms, enoki mushrooms, and shiitake mushrooms, which have high nutritional and economic value. However, the visual similarities between mushroom types complicate the manual identification process, which requires special expertise and is prone to errors. To overcome this, this study proposes a classification model for edible mushrooms using the Random Forest algorithm with a combination of Hue, Saturation, Value (HSV) color feature extraction and the Histogram of Oriented Gradients (HOG) shape. The dataset consists of 500 mushroom images from primary and secondary sources that are processed through augmentation, resizing, color conversion, and feature merging. The model is trained with three data sharing schemes (80:20, 70:30, and 60:40) and optimized using GridSearchCV. Evaluation based on accuracy, precision, recall, and F1-score shows that the 80:20 scheme produces the best performance with an accuracy of 97%. The optimal model was then implemented in a Python-based Graphical User Interface (GUI) capable of identifying fungal species quickly, accurately, and easily. This research contributes to the development of an efficient and reliable digital image-based fungal classification system.