Article Details

Abstract

To enhance the accuracy of customer value identification in the tea e-commerce industry, a dual-channel customer value identification method integrating a Large Language Model and a CART decision tree is proposed. Multi-source data such as orders, behaviors, conversations, and comments are integrated. The LLM is used to extract text semantic features, and the CART algorithm is combined with process structured behavioral indicators, thereby constructing a customer value classification and conversion prediction model. The model is verified using real data from 15 800 users of a tea flagship store. The integrated model achieves improvements of 12.3%, 9.8%, and 11.2% in accuracy, recall rate, and F1 score compares with the single CART model, respectively. Key influencing factors such as "frequency of tea culture content interaction" are also identified. This method combines the advantages of semantic understanding and interpretable classification, providing an effective solution for intelligent marketing in the tea e-commerce industry.

Keywords

Large Language Model; CART Decision Tree; customer identification; conversion prediction

References

1. [1] Te P，Alican F，Yin A O，et al. GPT-HTree: A Decision Tree Framework Integrating Hierarchical Clustering and Large Language Models for Explainable Classification [PP/OL].arXiv:2501.06981 [cs.CY].https://arxiv.org/abs/2501.06981.
2. [2] Liu X X，Chen X W. Who Decides: The Consumer or the Retailer? An LLM-Assisted Bayesian Framework for Modeling Purchase Decisions with Retailer Influence [J].Journal of Retailing and Consumer Services，2025，83：104201.
3. [3] Sadeghi A. Predicting Long-term Payment Risks in Energy Customers Using Large Language Models and BERTopic on Initial Conversational Data [D].Padova：Università di Padova，2024.
4. [4] Ariwiati，Susanty A，Batu K L. An Integrated LRFM and CLV-Based Customer Segmentation Model for B2B E-Marketplace Platform [C]//Proceedings of the 2025 IEEE International Conference on Information Technology and Electrical Engineering，2025：1-6.
5. [5] Shi W，Hu J T. Research on Precision Marketing of Cultural and Creative Agricultural Products E-Commerce in Big Data Era: An Analysis Based on User Profiles and LDA Topic Models [J/OL].SSRN Electronic Journal，2024.https://doi.org/10.2139/ssrn.4981129.
6. [6] Arjunan T，Umamageswari A. Five-Tier BI Architecture with Tuned Decision Trees for E-Commerce Prediction [J].Indonesian Journal of Electrical Engineering and Computer Science，2025，37（1）：446-456.
7. [7] Zheng Fu, He Lutong, Hong Linghui, etc Research on Consumer Satisfaction with Tea Products on E-commerce Platforms Based on Text Mining: A Case Study of Online Reviews of Kuding Tea on JD.com [J]. Business Development Economics, 2024 (15): 42-45
8. [8] Li Xiang Construction and Analysis of Product Selection Decision Model for Tea Social E-commerce Platforms [J]. Fujian Tea, 2025, 47 (2): 112-114
9. [9] Yang Zhiqi Analysis of Consumer Satisfaction with Yunnan Special Tea Brand E-commerce Based on Online Comments [J]. China Tea, 2023, 45 (8): 32-38
10. [10] Jiang M. A Hybrid Deep Clustering Approach for Cross-Border Live Streaming Precision Marketing [C]//Proceedings of the 2025 IEEE International Conference on Network, Multimedia and Information Technology，2025：1-5.

Funding

Jiangsu Province Modern Agricultural Vocational Education Industry Guidance Committee (Nonyhzwz202303); Research Project of Jiangsu Agricultural and Animal Husbandry Technology Vocational College (NSF2025RK03)

About the authors

Hong Wei (1980-), male, Han ethnicity, from Yangzhong, Jiangsu, associate professor, master's student, research interests include big data analytics and applications; Zhu Xiaoman (1995-), female, Han ethnicity, from Taizhou, Jiangsu, teaching assistant, master's student, research interests include big data analytics and applications.