| Summary: | Oil palm has become a commodity of global strategic importance due to its rapid expansion. Palm oil is widely utilised in food and as a biodiesel precursor. The oil boosts several countries’ economies, especially Malaysia’s. However, Ganoderma boninense causes basal stem rot (BSR), the most severe disease of oil palms. BSR management controls remain to be ineffective at the moment. There is currently no cure for BSR disease, and the only practical option is to extend the life of the oil palm tree. Thus, we demonstrate how a thermal image technique can be used to distinguish between healthy and BSR-infected trees. We assessed the feasibility of using WEKA standard machine learning algorithms (ML) such as Naive Bayes (NB), multilayer perceptron (MLP), and random forest (RF) to classify healthy and BSR-infected trees. Additionally, we emphasise the data imbalance technique in this study because, in reality, the number of healthy and BSR-infected is not uniform. Therefore, imbalanced data approaches such as random under-sampling (RUS), random over-sampling (ROS), and synthetic minority oversampling (SMOTE) are employed in this classification. In order to evaluate and compare various algorithms and imbalanced approaches, we described the receiver operating characteristic (ROC) curve region (AUC), the precision-recall curve (PRC), and the confusion matrix as an alternative in terms of the success rate of the non-infected and BSR-infected tree. We expect that our technique will assist non-expert users in identifying appropriate machine learning algorithms, resulting in improved performance for accurately predicting BSR disease. © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023.
|
|---|