HYBRID METAHEURISTIC FEATURE SELECTION FOR PLANT LEAF DISEASE PREDICTION USING DATA MINING TECHNIQUES

Abstract

The rapid proliferation of plant diseases due to climate change and scarcity of manpower calls for intelligent, automatic and efficient diagnostic systems that can aid in real-time decision making in agriculture. Image based data mining and machine learning have been found to be quite effective in solving such problems; however, the traditional feature selection methods fail to meet the requirements due to the difficulties with high-dimensional leaf images, redundant features and poor generalization across different kinds of crops and diseases. This paper introduces a Hybrid Metaheuristic Feature Selection Framework which uses a combination of several optimization techniques in a cooperative manner for enhancing the efficiency, stability and classification accuracy of plant leaf diseases. The hybridization of GA-PSO approaches, γ-ABC, Bat algorithm, Firefly-DE, PSO Stability, AFSO-SVM, FOA, Binary ALO, Binary WOA, and GWO are chosen to form this framework on the basis of previous research in this field and it makes use of the advantages of both global and local searches in identifying small and informative feature subsets despite the shortcomings such as slow convergence, parameter sensitivity and local optima issues.It is the intention of the hybrid model described herein to seamlessly integrate with the deep learning-based method of feature extraction, thus providing excellent results even in noisy and highly variable agricultural environments. Literature shows that hybrid metaheuristic models have proven superior to single algorithms in terms of accuracy, precision, recall, and computational time. Leveraging from this insight, the current research proposes the development of an integrated and scalable hybrid feature selection model, with an eye to enhancing the predictive power of plant diseases.

Plant phenotyping relevance

葉画像から植物病害を推定する特徴選択フレームワークの開発が研究の中心であり、植物の病害状態を直接評価する計算的フェノタイピング手法に該当する。

Code and data availability

公開本文の所在を確認できませんでした。非公開または購読が必要な可能性があります。

No evidence-backed public reproduction asset is currently recorded.