Rohde & Schwarz uses machine learning for network call stability assessment

Munich – German electronics firm Rohde & Schwarz has introduced a new machine learning-based method for assessing quality in mobile networks. The new "Call Stability Score" (CSS) feature aims to address challenges in traditional call drop ratio (CDR) measurement methods.

Traditionally, measuring CDR requires thousands of calls to achieve statistical significance, making it impractical for confined areas like shopping malls. Rohde & Schwarz's machine learning model analyzes numerous network parameters and scores each call based on its stability, irrespective of whether it dropped or not.

The "Call Stability Score" utilizes deep learning and recurrent neural networks, such as LSTM cells, to process time-series data. The model distinguishes between stable and dropped calls by learning a boundary within a high-dimensional feature space. This provides a more reliable assessment of call quality and overall network stability.

The company states that the new technology assists operators in gaining more precise insights into network performance and enhances the reliability of measurement campaigns. It also enables a more granular analysis of network quality across diverse conditions.