Machine Learning-Based Image Classification for Fish Disease Detection Using SVM and Random Forest

The fisheries industry in Indonesia's archipelagic regions faces serious challenges in achieving rapid and accurate fish disease detection.Conventional methods based on manual visual inspection are subjective, slow, and inefficient for large-scale industrial operations.This study proposes an image-based fish disease detection system using computationally efficient machine learning algorithms-Support Vector Machine (SVM) and Random Forest (RF).The dataset consists of 365 training images obtained through web scraping and 54 testing images collected primarily from Mardika Market, Ambon.The research stages include image preprocessing, feature extraction (color, texture, and shape), model training, and performance evaluation using metrics such as Accuracy, Precision, Recall, and F1-Score.The results show that in the validation stage, SVM performed better in the hold-out method with an accuracy of 79.45%, whereas RF demonstrated higher stability in 10-Fold Cross-Validation (average accuracy of 75.10%).Final testing on primary data indicated that RF achieved the best performance with an accuracy of 83.33% and an AUC of 0.91, surpassing SVM (77.78%;AUC 0.90).These findings confirm that the RF algorithm is more robust in handling real-world data variability.The hybrid approach-combining global and local data-proved effective in producing practically relevant models that can be implemented to support the sustainability of the fisheries industry in Indonesia's archipelagic regions.