A Hybrid Approach to Crop Disease Prediction: Combining Environmental and Image Data for Enhanced Accuracy

This study introduces a comprehensive methodology for predicting crop diseases by integrating Machine Learning (ML) techniques with Deep Learning (DL) models, aimed at aiding farmers in the early detection of plant diseases and optimizing crop selection. The proposed system utilizes a Random Forest Classifier for crop recommendations, taking into account essential agricultural factors such as nitrogen, phosphorus, potassium (NPK) levels, soil pH, temperature, humidity, and rainfall. Concurrently, MobileNetV2-based Convolutional Neural Networks (CNNs) are utilized for the identification of plant diseases through image classification, allowing for precise diagnosis of various crop diseases at their initial stages. The system also incorporates historical disease data alongside real-time weather information to enhance prediction accuracy and provide proactive management suggestions. The model undergoes training and validation using the Crop Recommendation Dataset and the PlantVillage Dataset to ensure its effectiveness across a range of crop types and environmental conditions. A user-friendly web interface was developed to enable straightforward image uploads, disease notifications, and actionable recommendations, with features supporting multilingual access and offline use. Experimental findings indicate a high level of accuracy in both crop recommendation and disease classification tasks, highlighting the system’s potential to promote sustainable agricultural practices. Future developments will focus on the integration of IoT sensors, GIS-based disease mapping, and real-time automated advisory services to further enhance predictive capabilities and bolster precision agriculture efforts.