Abstract

Telecommunications systems have evolved to include an ever-growing number of interdependent hardware and software components with complex interactions. This exponential increase in complexity affects the reliability and stability of network systems. This thesis provides two systematic approaches to improve the speed and quality of the Root Cause Analysis task in telecommunications systems.

The first approach introduces a new fault analysis framework based on association rule mining and evaluates it for telecommunication systems. The approach describes a strategy using association rules to specifically target faults while improving runtime performance relative to the standard Apache Spark implementation. It also introduces a novel filtering strategy called Cover Set filtering that prunes and merges rule sets to produce high-quality, concise and interpretable results. The proposed framework is evaluated with real-world telecommunication datasets. Compared with other strategies, we demonstrate a better rule diversity in general and a sufficiently compact fault analysis.

The second approach tackles Root Cause Analysis from the causal perspective. It is based on Counterfactuals and Nearest Neighbour Matching concepts to identify fault types and highlight the most fault contributing variables. The proposed framework is a proof of concept for finding the root cause of problems based on the causal learning technique. It is demonstrated to be highly compatible with numerical data and highly robust with noisy data.

In conclusion, the proposed frameworks improve the quality and performance of fault troubleshooting tasks in telecommunication systems. Last but not least, the proposed frameworks can be adapted to other information systems with minor modifications.