Combination of extraction features in detecting disease in corn stalks using gradient boosting method

Corn is one of the essential staple crops. It is a plant for many communities around the world. Since it plays a crucial role in global food security such as human survival, therefore corn production must be carefully managed to avoid negative impacts. It can be a significant issue if the selection of high-quality corn kernels cannot be performed. If plants are infected, they may not grow well, produce small cobs, or even die before harvest. Consequently, this results in reduced crop yields, which can harm farmers and diminish food supplies. The preventive measures can be proposed by leveraging technological advancements through data mining. It covers building a specialized system for agriculture to assist farmers in monitoring, diagnosing, and managing diseases in corn by utilizing images and artificial intelligence to identify diseases based on images of the corn stalks. This research proposes machine learning approaches, such as Artificial Neural Networks, Gradient Boosting, and AdaBoost in classifying corn stalk diseases. In addition, a combination of LBP (Local Binary Pattern) and HSV (Hue Saturation Value) feature extraction is used in this research. The dataset used is digital images from 750 records into 5 data classes (Healthy, Erwina Carotovora, Pythium, Stenocarpella, and Gibberella). The results of this study, gradient boosting is the best method which has both an accuracy of 81.5% and an AUC score of 96.3%.