Intelligent Telecom Solutions: Employing Large Language Models to Address Base Station Anomalies and Alarm Prioritization

Telecom network issues disrupt critical services and lead to substantial financial losses. This study focuses on two key tasks: generating solutions for Base Station (BS) problems and prioritizing BS alarms using Large Language Models (LLMs). The application of these models improves telecom network management by reducing downtime and increasing reliability during emergencies. Six LLMs were evaluated: Mixtral, Llama, and Gemini for generating solutions. DistilBERT, DistilGPT-2, and ELECTRA for alarm prioritization. Mixtral achieved a high cosine similarity score of 0.9647, outperforming Llama, Gemini, and traditional models such as Markov in resolving BS anomalies. DistilBERT demonstrated superior accuracy, precision, recall, and F1 score in classifying alarm priorities for BS compared to DistilGPT-2, ELECTRA, and conventional models such as SVM and Naive Bayes. These findings highlight the potential of LLMs to significantly improve stability and reliability in the telecommunications industry, providing valuable advances over traditional methods.