FREQUENCY BASED RANKING SUPPORT VECTOR MACHINE (FRSVM) ALGORITHM FOR FUNGAL DISEASE DETECTION IN WHEAT CROP BASED ON KCC DATASET

Abstract

Wheat (Triticum Aestivum) is one of the most significant rabi (winter) crops in India, serving as a staple food for over 40% of the population. It is the second most-produced cereal after rice, contributing significantly to India’s food security and agricultural economy. Early detection of fungal diseases in wheat is critical for preventing yield loss and ensuring effective crop management. Farmers frequently describe crop symptoms through natural language queries submitted to agricultural advisory systems such as call centers or digital platforms. Manually analyzing these queries is time-consuming and inefficient when dealing with large volumes of farmer requests. This research proposes a Frequency based Ranking Support Vector Machine (FRSVM) algorithm for automatic classification of wheat fungal disease symptoms collected from Kisan Call Center (KCC) farmer queries in ICAR dataset. The dataset contains symptom-based farmer queries covering five major fungal diseases of wheat: Head Scab, Karnal Bunt, Loose Smut, Powdery Mildew, and Pink Stem Borer. The proposed approach utilizes text preprocessing and TF-IDF feature extraction to convert farmer symptom queries into numerical representations. A custom Symptom Scaler emphasizes disease-specific keywords such as “rust,” “powder,” and “blight,” while the ranking score prioritizes vital disease based on their frequency. The proposed model achieved 93.5% accuracy, with other Machine Learning baseline algorithms including Naive Bayes, Random Forest, Linear SVM and XG Boost classifiers. The results demonstrate that frequency based ranking mechanism with machine learning provides an effective solution for automatic fungal disease identification from farmer queries, enable faster decision-making for wheat crop disease management and also significantly improves automated agricultural advisory systems.