Path loss modeling based on neural networks and ensemble method for future wireless networks

Area of Science:

• Wireless Communication
• Machine Learning
• Electromagnetics

Background:

• Technological advancements necessitate higher data speeds, driving demand for higher frequency bands (millimeter-wave and subterahertz).
• Existing path loss prediction models for 5G and beyond lack the flexibility and accuracy needed for challenging environments.
• Accurate path loss prediction is crucial for deploying wireless networks with guaranteed quality of service.

Purpose of the Study:

• To develop and evaluate advanced machine learning-based path loss prediction models for high-frequency bands.
• To compare the performance of Artificial Neural Network (ANN), Recurrent Neural Network-Long Short-Term Memory (RNN-LSTM), Convolutional Neural Network (CNN), and an ensemble-method-based model.
• To identify the most efficient and accurate model for path loss prediction in complex indoor environments.

Main Methods:

• Implementation of ANN, RNN-LSTM, and CNN models for path loss prediction.
• Development of a novel ensemble-method-based neural network path loss model.
• Performance analysis based on prediction accuracy, stability, feature contribution, and computational time.
• Training and testing using data from indoor corridor measurement campaigns (line-of-sight and non-line-of-sight).

Main Results:

• The ensemble-method-based model demonstrated superior prediction accuracy and efficiency compared to individual ANN, RNN-LSTM, and CNN models.
• The study provided an extensive performance analysis of all four models.
• The proposed ensemble model showed high prediction accuracy and efficiency in complex environments.

Conclusions:

• The ensemble-method-based path loss prediction model is a promising solution for high-frequency wireless communication.
• This model offers enhanced accuracy and efficiency, crucial for optimizing wireless network deployment.
• The findings support the use of advanced machine learning techniques for reliable path loss prediction in challenging scenarios.