AI-POWERED PLANT HEALTH ASSESSMENT: AUTOMATED CLASSIFICATION FOR ENHANCED CROP MONITORING AND PRODUCTIVITY

Abstract

Plant disease classification plays a vital role in advancing modern agriculture, transitioning from traditional manual diagnosis to intelligent, automated systems powered by machine learning. Historically, identification of plant diseases relied on visual inspections, expert advice, and lab tests— methods that were accurate for small-scale use but often subjective, slow, and inconsistent. These limitations resulted in delayed treatment and substantial crop losses, highlighting the inefficiency and high cost of conventional approaches, especially at scale. To address this, the proposed system introduces an innovative machine learning-based solution capable of accurately classifying plant diseases using sparse and categorical IoT data. It incorporates comprehensive data preprocessing techniques, including handling missing values, label encoding, and class imbalance correction using the Synthetic Minority Oversampling Technique (SMOTE), ensuring a high-quality dataset for model training. The classification pipeline integrates multiple models—Gaussian Naive Bayes, Support Vector Machines, K-Nearest Neighbors, and a novel Decision Tree Classifier. Among these, the Decision Tree model demonstrated superior performance, achieving an accuracy of 99.07% with precision, recall, and F1-scores consistently exceeding 98%, confirming its robustness and reliability. This research is significant in offering real-time, data-driven diagnostics that enable early disease detection and precise pesticide recommendations. It not only improves crop yield and reduces financial losses but also promotes environmentally sustainable agriculture by limiting excessive chemical usage. By overcoming the limitations of traditional methods—such as subjectivity, delay, and lack of scalability—this system presents a transformative approach to plant disease management through advanced machine learning, marking a pivotal shift toward precision agriculture.

Code and data availability

The article describes a plant disease classification pipeline (GNB, SVM, KNN, DTC on IoT data with SMOTE) but contains no public dataset, image/sensor data deposit, author code repository, trained model release, or availability statement. Figures are GUI screenshots and plots; references are cited prior work. No paper-

No evidence-backed public reproduction asset is currently recorded.