Article Details

Abstract

To address the existing problems of insufficient accuracy of user portraits in intelligent advertising recommendation, high privacy risks, poor real-time delivery performance and low marketing efficiency for Small and Medium-sized Enterprises (SMEs), this paper constructs an intelligent advertising recommendation system for SMEs that integrates the fused LightGBM-KNN algorithm with privacy-preserving computation. Privacy-preserving computation is adopted to achieve secure fusion of multi-modal operator data. A lightweight user portrait model integrating LightGBM and the KNN algorithm is built to accurately classify residential and commercial users. Kafka is utilized to realize real-time secure backhaul of advertising data and iterative optimization of delivery strategies. Comparative experiments verify that the Click-Through Rates (CTR) of advertisements targeting residential and commercial users increase by 83.96% and 68.09%, respectively, and the overall conversion rate increases by 201.47%. The optimized CTR reaches 3.95%, and the conversion rate reaches 2.05%. Further performance improvements can be achieved through technological upgrades in future research. Balancing privacy compliance and advertising delivery efficiency, the proposed system provides effective technical support for the precision marketing of SMEs.

Keywords

privacy computing; fused LightGBM-KNN algorithm; user portrait; Kafka; intelligent advertising recommendation; Small and Medium-sized Enterprises (SMEs)

References

1. [1] Huo Wei, Yu Yu, Yang Ku, etc Research Progress and Application of Privacy Protection Computing Password Technology [J]. Chinese Science: Information Science, 2023, 53 (9): 1688-1733
2. [2] Qian Wenjun, Shen Qingni, Wu Pengfei, etc Research progress on privacy protection technology in big data computing environment [J]. Journal of Computer Science, 2022, 45 (4): 669-701
3. [3] Xie Junfei, Zhang Haiqing, Li Daiwei, etc Optimization of Deep Forest Algorithm Based on Lightgbm and XGBoost [J]. Journal of Nanjing University (Natural Science), 2023, 59 (5): 833-840
4. [4] Li Jiahua Personalized recommendation algorithm for artificial intelligence cross-border e-commerce guide platform based on big data [J]. Science, Technology and Engineering, 2019, 19 (14): 280-285
5. [5] Hou Li, Wang Jian User Big Data Periodic Intelligent Recommendation Algorithm Based on Collaborative Filtering [J]. Computer Simulation, 2023, 40 (3): 476-479489
6. [6] Qin Xuebing Research on the Evolution of Intelligent Advertising Industry from the Perspective of Innovation Ecosystem Theory [J]. Contemporary Communication, 2022 (2): 67-69
7. [7] Wang Xue, Cha Xianjin, Mei Xiao, etc Research on User Information Avoidance Behavior in Intelligent Recommendation of Mobile Social Media from the Perspective of Privacy Concerns [J]. Journal of Intelligence, 2023, 42 (5): 598-610
8. [8] Zhang En, Huang Yuchen, Zheng Dong, etc Threshold multi-party privacy set intersection protocol that balances communication rounds and computational overhead [J]. Journal of Software, 2026, 37 (4): 1819-1837
9. [9] Yang Shuai, Han Yiliang, Wu Xuguang, etc A large model bidirectional privacy authorization based on privacy set intersection [J]. Journal of Network and Information Security, 2025, 11 (5): 101-114
10. [10] Kang Xin, Zeng Yong, Wang Zhuzhu, etc Privacy set intersection scheme based on unintentional polynomial evaluation [J]. Chinese Science: Information Science, 2026, 56 (2): 362-377

About the authors

Liu Min (1975-), female, Han ethnicity, from Taizhou, Zhejiang, senior engineer, master's student, research interests include big data and artificial intelligence; Huang Chunming (1986-), male, Han ethnicity, from Lingao, Hainan, assistant engineer, bachelor's degree, research interests include big data and artificial intelligence; Chen Hui (1978-), male, Han ethnicity, from Xinjian, Jiangxi, engineer, master's student, research interests include business support system construction, software intelligence, big data applications and data governance.