An IoT-Enabled Hybrid Deep Q-Learning and Elman Neural Network Framework for Proactive Crop Healthcare in the Agriculture Sector

Abstract

Emerging sensing technology and the artificial intelligence (AI) has lifted the agriculture sector by offering crop health monitoring and enabling real-time decision making. However, the heterogeneous nature of IoT devices results in massive data with distinct features that present challenges for individual AI models to comprehend the inherited data pattern, thereby necessitating advanced models. Consequently, we introduce an IoT coupled hybrid framework that integrates Deep Q-Network and Elman Neural Network (ENN) for proactive crop healthcare in the agriculture sector. The developed hybrid framework utilizes the IoT system for crop monitoring data and incorporates ENN, which leverages the Recursive Pattern Elimination technique to evaluate the data patterns and extract the optimal pattern related to crop health. Subsequently, the developed framework utilizes Deep Q-Network to comprehend the inherited data pattern related to the crop health for informed decision-making purposes. The proposed hybrid framework is applied to publicly available Field and Greenhouse crop datasets collected through the IoT system and is validated against state-of-the-art models focused on crop healthcare. The results showed that the proposed ENN-DQN framework achieved a high accuracy of 99.77%, precision of 99.52%, recall of 99.93%, and F-score of 99.76%. Moreover, a detail of the DQN action distribution is presented, and the results are validated through robustness analysis against different levels of heterogeneity, statistical analysis with a 95% confidence interval, and computational complexity analysis. A source code for this study is openly accessible at: GitHub repository