Lightweight Feature Selection Method using Quantum Statistical Ranking and Hybrid Beetle-Bat Optimization for Smart Farming

Abstract

The advancement of IoT-enabled smart farming systems has generated massive high-dimensional datasets, creating challenges in feature selection, classification accuracy, and computational efficiency. Existing feature selection techniques, including ReliefF, LASSO, and Recursive Feature Elimination (RFE), achieve moderate performance but struggle with scalability and runtime constraints. Similarly, wrapper-based optimization methods like Genetic Algorithms (GA) and Particle Swarm Optimization (PSO) provide higher accuracy but incur significant computational overhead, making them unsuitable for real-time IoT applications. To address these limitations, this study proposes a Quantum-Enhanced Mutual Rank Index with Beetle-Bat Optimization (QStat-BBO) framework for lightweight and adaptive feature selection. The proposed approach integrates Quantum-Enhanced Mutual Rank Index (Q-MRI) to prioritize features based on mutual dependencies and utilizes Beetle-Bat Optimization (BBO) to refine optimal feature subsets efficiently. Three IoT-based agricultural datasets from smart farming environments are used to evaluate the framework. Experimental results demonstrate that QStat-BBO consistently outperforms state-of-the-art methods, achieving up to 97.4% classification accuracy, 0.975 F1-score, and an average feature reduction rate of 63.5%, while reducing runtime by nearly 40% compared to traditional metaheuristics. These results confirm the effectiveness of QStat-BBO in enhancing prediction performance, reducing redundancy, and improving computational efficiency, making it well-suited for resource-constrained IoT-based agricultural analytics.