A Multivariate Time-Series Based Approach for Quality Modeling in Wireless Networks

Area of Science:

• Telecommunications Engineering
• Network Performance Analysis
• Machine Learning Applications

Background:

• Accurate estimation of Key Quality Indicators (KQIs) is crucial for managing modern wireless networks.
• Existing methods may struggle with the dynamic and non-stationary nature of advanced wireless environments like 5G and 6G.
• The need for efficient, online KQI estimation methods is increasing with network complexity.

Purpose of the Study:

• To develop and evaluate a novel method for estimating KQIs from node measurements in wireless networks.
• To design a framework adaptable to 5G and 6G systems, capable of handling non-stationary conditions.
• To demonstrate the feasibility of the proposed method using real-world network data.

Main Methods:

• Utilized multivariate adaptive filtering and a clustering algorithm for KQI estimation.
• Generated KQI time-series data suitable for network management systems.
• Implemented a framework designed for fast, online training and flexibility.

Main Results:

• The proposed method successfully estimated KQIs from live heterogeneous network data.
• Initial results demonstrated the method's effectiveness compared to traditional linear regression techniques.
• The framework showed adaptability to non-stationary network conditions.

Conclusions:

• The presented method offers a viable approach for real-time KQI estimation in advanced wireless networks.
• The framework's design supports application in future 5G and 6G systems.
• Further research can refine the algorithms for enhanced performance and broader applicability.