Environmental risk assessment based on multiscale spatial recurrent neural network algorithm for IoT agriculture area

Area of Science:

• Agricultural Technology
• Data Science
• Network Engineering

Background:

• Smart agriculture utilizes Internet of Things (IoT) devices and sensors to monitor environmental conditions, aiming to boost crop yields and optimize resource management.
• High data traffic in IoT systems poses a challenge, leading to delays in accessing critical information and hindering system performance.
• Current traffic analysis methods often overlook the marginal rate of traffic features, impacting the efficiency of smart agricultural environments.

Purpose of the Study:

• To propose and evaluate a novel Multiscale Spatial Recurrent Neural Network (MSRNNet) for classifying Internet of Things (IoT) traffic within smart agricultural systems.
• To address the challenge of redundant traffic in IoT data collection by developing an effective feature selection and classification approach.
• To enhance the performance and reliability of smart agriculture systems through optimized IoT traffic management.

Main Methods:

• Data preprocessing using Box-Plot Normalization (BPN).
• Feature evaluation and selection using Exhaustive Traffic Information Rate (ETIR) and AntLion Behavior Optimization (ALBO) algorithms for dimensionality reduction.
• Classification of the optimized IoT traffic data using the proposed Multiscale Spatial Recurrent Neural Network (MSRNNet).

Main Results:

• The MSRNNet model achieved high performance metrics: 97.08% accuracy, 96.05% precision, 94.25% recall, and 95.71% F1-score.
• The proposed method demonstrated a low misclassification rate of 1.25%.
• The system exhibited efficient processing with a time complexity of 85.49 ms, confirming its reliability.

Conclusions:

• The developed MSRNNet model effectively classifies IoT traffic, offering a significant improvement for smart agriculture systems.
• The combination of BPN, ETIR, and ALBO provides an efficient method for data preprocessing and feature selection, reducing traffic complexity.
• The research validates the proposed approach as a reliable and effective solution for enhancing efficiency and productivity in smart farming through optimized IoT data management.