Enhanced Plant Leaf Disease Detection Using Modified Logistic Regression for Sustainable Agriculture Practices

Pepper, Hibiscus, and Basil are medicinal plants with a rich history in traditional medicine and health benefits. They are essential in culinary and medicinal applications, contributing to natural health solutions. Disease detection is crucial to protect their agricultural, economic, and medicinal value. Early detection minimizes crop losses, maintains plant health, and ensures plant availability for traditional medicine and culinary uses. This promotes sustainable and eco-friendly agricultural practices. Traditional logistic regression for plant leaf disease detection struggles with imbalanced data and a fixed linear decision boundary, making it less effective in capturing complex disease patterns. The modified logistic regression model with the One Half Constant improves performance metrics and handling intricate features of leaf images by addressing class imbalance more effectively. It adjusts the decision boundary to handle imbalanced datasets, enhancing classification accuracy for minority classes while maintaining simplicity and interpretability. This study uses a dataset collected from Kaggle and surrounding of Kadapa district AP, India. For the evaluation of the proposed model in disease detection, the traditional logistic regression and other machine learning algorithms were used, and the corresponding key metrics of accuracy, precision, recall, false positive rate (FPR) and F-Measure were assessed. A comparison with existing methods show overwhelming improvement of 32.94% in accuracy, 17.64% in precision, 33.6% in recall, 86.91% in FPR improvement, 34.6% in F-Measure. The proposed approach seeks to improve overall diagnostic accuracy, thereby providing a reliable tool for early detection and treatment planning in clinical sectors.