ABSTRACT
This paper focuses on the use of Deep Belief Networks and Isolation Forests as the methods to diagnose aquatic plant diseases . Ten experiment trials were conducted, which showed that DBNs had an average accuracy of 86% and an average precision, recall, and F1-score of 88%, 84%, and 86%, respectively. On the other hand, Isolation Forests demonstrated high true positive rates of 91% on average and unsatisfactory false positive rates of 4% on average. Moreover, the average area under the ROC curve was 0.93, which indicates a high discrimination ability between positive and negative instances. The findings show that machine learning techniques can be used to improve the diagnosis of diseases in aquatic plants, which is vital for implementing prevention or control measures to save ecosystems. The research results documented in this paper reveal that the use of DBNs and Isolation Forests power increases the diagnostic accuracy, enabling efficient management actions to counter environmental disruptions. This investigation contributes to the development of environmental monitoring and highlights that versatile tools could be implemented to address and manage the ecological health of aquatic plants.
