DAMPSO: Dynamic Accelerated Memory-Based PSO for Hyperparameter Tuning of Plant Disease Classifiers

Plant disorders cause substantial threats to agriculture, impacting crop conditions and output. In recent years, the recognition and utilization of machine learning have grown substantially as a valuable asset, particularly in automating the recognition and categorization of foliage diseases. This advancement has resulted in enhanced agricultural practices. However, traditional detection methods suffer from drawbacks such as high costs, error-proneness, inconsistency, time-consuming procedures, and the need for enhanced efficiency. To overcome these constraints, this paper introduces a model that utilizes imaging techniques to detect and categorize foliar diseases in maize and cotton plants. The study utilizes a dataset comprising 2543 images of maize leaves and 1443 images of cotton leaves with distinct disease classes sourced from online platforms. The study incorporates hyperparameter tuning of the aforementioned classifiers using machine learning-based hyperparameter tuning techniques such as random search, grid search, and nature-inspired algorithms. In particular, a novel algorithm named Dynamic Accelerated Memory-Based PSO (DAMPSO) is proposed for hyperparameter tuning. Experimental analysis reveals that DAMPSO surpasses other tuning techniques, yielding the highest accuracy of 95% when applied to the XGBoost classifier.