Research on Optimization of Large Model Recommendation Algorithm for Intelligent Customer Service

DOI: https://doi.org/10.62517/jike.202604120

Author(s)

Yan Yang, Sai Wang

Affiliation(s)

Computer School, Central China Normal University, Wuhan, China

Abstract

This paper addresses the bottlenecks of large language models in intelligent customer service scenarios, such as high response latency, significant resource consumption, and lack of domain knowledge, by conducting systematic optimization research on recommendation algorithms. Model lightweighting is achieved through the integration of model pruning and knowledge distillation techniques. Training efficiency is enhanced by combining distributed and mixed-precision training. Furthermore, domain knowledge graphs are innovatively introduced to improve the accuracy and reliability of generated responses. Experiments demonstrate that the optimized system significantly accelerates response times while effectively improving accuracy in handling complex specialized queries, providing a viable pathway for efficient and precise application of large models in vertical business scenarios.

Keywords

Recommendation Algorithm; Large Language Model; Intelligent Customer Service

References

[1] YUAN P, ZHANG F. Application of Adaptive Large Model Architecture in Intelligent Recommendation [J]. Information Recording Materials, 2025, 26 (09): 95-97. [2] WU Y, LU J. Multi-modal Information Generation and Recommendation Driven by Large Models [J]. Journal of Henan Normal University (Natural Science Edition), 2025, 53 (05): 145-151+181. [3] NIU Y, HAO B, ZHAO Z. Research on the Construction and Practice of Personalized Resource Recommendation System for University Libraries Based on Large Models [J]. New Century Library, 2025, (07):66-73. [4] YOU X, LI S, SHAO H. Construction of an Intelligent Question and Answer System for Clothing Recommendation Based on Large Models [J]. Woolen Textile Technology, 2025, 53 (05): 87-94. [5] YANG Y, PAN S, LIU X, et al. Multi-modal Knowledge Graph and Collaborative Decision-Making of Large Models for Risk Management in Hydraulic Engineering [J]. Journal of Water Resources, 2025, 56 (04): 519-530. [6] ZHOU X. Research on Efficient Recommendation Algorithm Based on Large Language Model [D]. University of Electronic Science and Technology of China, 2025. [7] LIU P, ZHANG M, WANG P, et al. Algorithm Optimization and Performance Evaluation of Intelligent Knowledge Recommendation System for Convenience Hotline Work Orders Based on Large Models [J]. Digital Technology and Applications, 2025, 43 (03): 16-18. [8] WANG M, GAO X, WANG S, et al. Research on Recommendation System Based on Knowledge Graph and Large Language Model Enhancement [J]. Big Data, 2025, 11 (02): 29-46. [9] MA X, GAO J, LIU Y, et al. Construction of a Customer Service Knowledge Recommendation Model Driven by Intent Understanding [J]. Journal of South China University of Technology (Natural Science Edition), 2025, 53 (03): 40-49. [10] ZHANG Y. Research on the Construction and Recommendation of Course Content Knowledge Graph Based on Large Language Model in Smart Education [D]. Sichuan Normal University, 2024. [11] ZHANG X, ZHANG L, YAN S, et al. Personalized Learning Recommendation Based on Knowledge Graph and Large Language Model Collaboration [J]. Computer Applications, 2025, 45 (03): 773-784. [12] ZHU M. Research on Personalized Resource Recommendation Method Based on Large Language Model [J]. Journal of Lanzhou University of Arts and Sciences (Natural Science Edition), 2024, 38 (05): 59-64. [13] ZHOU X, DENG X, HUANG W. Large Models and Recommendation Systems Open a New Chapter in Personalized Recommendation [J]. Shanghai Informatization, 2024, (09):35-38. [14] WU G, QIN H, HU Q, et al. Research on Large Language Models and Personalized Recommendations [J]. Journal of Intelligent Systems, 2024, 19 (06): 1351-1365. [15] LIU L. Research and Application of Professional Recommendation Knowledge Graph Construction Technology Based on Large Language Model [D]. Hangzhou University of Electronic Science and Technology, 2024.