Disease Progress Forecasting System Using EfficientNet and LSTM

Modern agriculture leverages various smart farming technologies to manage crops efficiently, but ensuring optimal growth for each crop remains challenging. This paper proposes a novel approach that integrates an image classification model with a time series analysis model to address this issue. The proposed system uses EfficientNet to classify crop diseases and LSTM (Long Short-Term Memory) to analyze time series data to predict the progression of the disease based on its presence. The model performance was evaluated using Accuracy, Precision, Recall, and F1 Score metrics. The image classification model achieved 91% Precision, 92% Recall, 97% Accuracy, and 94% F1 Score, outperforming ResNet, DenseNet, and SENet. Additionally, the model&quots reliability was verified using Grad-CAM (Gradient-weighted Class Activation Map), an XAI technique. The time series analysis model demonstrated a Recall of 88% and Precision and F1 Score of 86%. The proposed system is accessible via a web page, allowing easy access for users, and includes features for writing posts and comments to facilitate a user community. This system aims to advance smart farming technology, increase crop yields, and minimize agricultural losses, ultimately contributing to improved crop management efficiency.