Classification of Non-Infected and Infected with Basal Stem Rot Disease Using Thermal Images and Imbalanced Data Approach

Basal stem rot (BSR) disease occurs due to the most aggressive and threatening fungal attack of oil palm plant known as Ganoderma boninense (G. boninense). BSR is a that has significant impact on crops in Malaysia Indonesia. Currently, only sustainable strategy available extend life trees, there no effective treatment for disease. This study used thermal imagery identify features classify non-infected BSR-infected trees. The aims this were (1) potential temperature (2) examine performance machine learning (ML) classifiers (naïve Bayes (NB), multilayer perceptron (MLP), random forest (RF) trees are BSR-infected. sample size consisted 55 uninfected 37 infected We imbalance data approaches such undersampling (RUS), oversampling (ROS) synthetic minority (SMOTE) these classifications different sizes. found Tmax feature beneficial characteristic classifying or Meanwhile, ROS approach improves curve region (AUC) PRC results compared single approach. result showed combination TmaxTmin had higher correct classification G. ROS-RF robust success rate, correctly 87.10% 100% by boninense. In terms model using variables, Tmax, an excellent receiver operating characteristics (ROC) 0.921, precision–recall (PRC) gave value 0.902. Therefore, it can be concluded ROS-RF, predict with relatively high accuracy.