Quality of Service Prediction via Large Language Models

Huiying Liu, Zekun Zhang, Lei Sang, Qilin Wu, Yiwen Zhang

Abstract: Large language models (LLMs) have been the center of many research areas in recent years, and have a wide range of application scenarios from text understanding and content creation, to data analysis and human-computer interaction. After being trained on trillions of text tokens, LLMs exhibit excellent generalization abilities, and can work on various tasks with minimum alteration or finetuning. Such capability is potentially useful for the task of service recommendation, where the sparsity of training data has hindered the performance of many existing algorithms significantly. In this paper, we explore the possibility of using LLMs to predict the quality of service (QoS) value directly for a pair of web user and service. We propose the Large Language Model for Service Recommendation (LLMSRec), a framework that treats each web user and service as a descriptive natural language sentence, and learns to understand the QoSrelated features of them from limited historical invocation data. On the WSDream dataset, LLMSRec can make accurate predictions under high sparsity levels, and outperforms QoS prediction baseline methods consistently.

External IDs:dblp:conf/icws/LiuZSWZ25