Comprehensive analysis of mango leaf disease prediction using traditional machine learning

Abstract

Mangoes are grown in different countries for economic value, but tree’s leaves are prone to getting certain diseases that may be destructive on the economical crops. This paper therefore presents in this research the use of machine learning to provide a novel approach towards the identification and categorization of diseases on mango leaves. Primarily, the use of image analysis and feature extraction enables us to perceive the pictorial characteristics associated with all the diseases using the Histogram of Oriented Gradients (HOG) features. Combating challenges such as the inequity of the mango leaf images’ feature representation, distribution imbalance between classes, and model optimization, the proposed study employs a dataset with 4,000 images collected from multiple plantations. The results obtained by the experiment show the works of Logistic Regression, SVM, Random Forest, and XGBoost algorithms in distinguishing the healthy and the malignant mango leaves more accurately. Also, the implications of this research on supporting the sustainability of agriculture are discussed, especially regarding the effectiveness of the proposed classification models for early diagnosis and mitigation of diseases in mango production. Altogether, this research study is significant to the development of the methodologies within the field of precision agriculture, as well as stress the necessity of using the machine learning approach to the problem of crop disease detection.