Evaluation of Deep Learning Methods in a Dual Prediction Scheme to Reduce Transmission Data in a WSN

Area of Science:

• Computer Science
• Electrical Engineering
• Network Engineering

Background:

• Wireless Sensor Networks (WSN) face significant challenges in extending the operational lifetime of battery-powered sensors.
• Reducing energy consumption is paramount for prolonging sensor network functionality.
• Data prediction offers a viable strategy to minimize data transmission, thereby conserving energy.

Purpose of the Study:

• To compare the efficacy of various deep learning methods for data prediction in WSNs.
• To evaluate the impact of these methods on reducing data transmission and energy consumption.
• To identify the optimal deep learning model for enhancing sensor lifetime.

Main Methods:

• Implementation and comparison of different deep learning model architectures for a dual prediction scheme.
• Detailed presentation of model structures and parameters.
• Evaluation using diverse performance metrics, including data transmission reduction and data accuracy at the Base Station (BS).

Main Results:

• The deep learning model incorporating Attention demonstrated superior performance.
• This Attention-based model significantly reduced the amount of data requiring transmission.
• The model successfully maintained a high degree of accuracy, closely representing the originally measured data.

Conclusions:

• Deep learning, specifically models with Attention mechanisms, presents a highly effective approach for data prediction in WSNs.
• Utilizing Attention-based models can lead to substantial energy savings by minimizing data transmission.
• These findings highlight a promising direction for extending the lifetime of wireless sensor networks.