Optimization of Machine Learning Model for Sugarcane Leaf Detection Using Ensemble Methods

Abstrak

Sugarcane leaf identification plays an important role in supporting agricultural monitoring and crop management. However, variations in leaf appearance and background conditions often reduce the performance of single classification models. This study aims to analyse and compare the performance of Support Vector Machine (SVM), AdaBoost, and ensemble models that combine SVM and AdaBoost for sugarcane leaf image classification. The data used in this study was obtained from a public dataset available on Kaggle and modified through image selection with a total of 1,391 data points grouped into two classes, namely sugarcane (587 image data points) and non-sugarcane (804 image data points), as well as pre-processing steps. Feature extraction was performed to represent leaf characteristics prior to classification. The models were evaluated using accuracy, precision, recall, and F1-score metrics. The experimental results showed that the SVM-AdaBoost ensemble model achieved the best performance among all models tested, with an accuracy value of 93.55% and an F1-score of 93.47%, demonstrating its effectiveness in improving classification reliability. These findings indicate that ensemble learning can improve classification performance for sugarcane leaf images and can be considered as an alternative approach for agricultural image analysis applications.