顧客終身價值(LTV)預測是現代行銷與客戶管理的核心工具,能協助企業優化行銷策略、提升客戶留存與利潤。隨著數據規模擴大與消費行為日益多變,LTV預測面臨資料稀疏、分布偏態與不確定性等多重挑戰,推動了多種創新預測方法的發展。
主流預測方法與技術演進
- 傳統統計與機器學習方法:早期方法如RFM模型、BG/NBD、隨機森林與XGBoost等,適用於有明確交易記錄的場景,但對於資料稀疏或長尾分布的LTV預測效果有限 (Su et al., 2023; Ejgerdi & Kazerooni, 2023; Kanchanapoom & Chongwatpol, 2023; Chamberlain et al., 2017)。
- 深度學習與概率模型:深度神經網絡(DNN)、零膨脹對數常態(ZILN)分布、貝葉斯模型等,能同時處理用戶流失概率與極端高值,提升預測準確性與不確定性量化 (Wang et al., 2019; Liu et al., 2024; Cao et al., 2024; Calabourdin & Aksenov, 2023)。
- 多視角與多源學習:多視角對比學習、跨域遷移學習、分布自適應網絡等,針對資料稀疏、分布多樣與跨平台應用,顯著提升模型魯棒性與泛化能力 (Wu et al., 2023; Li et al., 2022; Zhou et al., 2024; Pan et al., 2025; Wu et al., 2023; Xing et al., 2021)。
實務挑戰與解決方案
- 資料稀疏與長尾分布:多數用戶貢獻低,少數用戶貢獻極高,需用分布自適應、多專家模型、兩階段預測等方法處理 (Li et al., 2022; Zhou et al., 2024; Pan et al., 2025; Liu et al., 2024; Zhang et al., 2024)。
- 特徵缺失與不確定性:針對特徵缺失,採用特徵感知路由融合網絡、蒙地卡羅Dropout等技術,提升預測穩健性與信心度量 (Cao et al., 2024; Yang et al., 2023)。
- 跨域與多源數據融合:跨平台、跨場景遷移學習(如CDLtvS、CDAF),有效利用上游豐富數據提升下游預測表現 (Su et al., 2023; Zhou et al., 2024; Pan et al., 2025)。
LTV預測方法比較
| 方法/模型 | 處理稀疏/長尾 | 不確定性量化 | 跨域適應 | 實務應用成效 | 代表文獻 |
|---|---|---|---|---|---|
| 傳統機器學習 | 一般 | 否 | 否 | 中等 | (Ejgerdi & Kazerooni, 2023; Chamberlain et al., 2017) |
| 深度概率模型 | 良好 | 是 | 否 | 優秀 | (Wang et al., 2019; Liu et al., 2024) |
| 多視角/多源學習 | 優秀 | 是 | 是 | 優秀 | (Wu et al., 2023; Li et al., 2022; Zhou et al., 2024; Pan et al., 2025) |
| 特徵感知/不確定性 | 良好 | 是 | 否 | 優秀 | (Cao et al., 2024; Yang et al., 2023) |
Figure 1: LTV 預測方法與應用比較
主要應用場景與成效
- 遊戲、電商、廣告等產業已大規模部署LTV預測系統,帶來顯著ROI提升 (Wu et al., 2023; Li et al., 2022; Wang et al., 2019; Pan et al., 2025; Liu et al., 2024; Yan & Resnick, 2023)。
- 推薦系統領域正從短期點擊優化轉向長期LTV最大化 (Wu et al., 2023; Xing et al., 2021)。
- 若需深入特定方法或產業案例,可參考 (Wu et al., 2023; Li et al., 2022; Wang et al., 2019; Zhou et al., 2024; Pan et al., 2025; Liu et al., 2024; Cao et al., 2024; Chamberlain et al., 2017; Yan & Resnick, 2023; Xing et al., 2021)等論文。
總結
LTV預測技術正快速演進,從傳統統計到深度學習、跨域遷移與不確定性建模,能有效應對資料稀疏、分布偏態與多源融合等挑戰。未來發展將聚焦於提升模型泛化能力、解釋性與實時應用價值,助力企業精準決策與長期成長。
References
Wu, C., Li, J., Jia, Q., Zhu, H., Fang, Y., & Tang, R. (2023). Contrastive Multi-view Framework for Customer Lifetime Value Prediction. ArXiv, abs/2306.14400. https://doi.org/10.48550/arXiv.2306.14400
Li, K., Shao, G., Yang, N., Fang, X., & Song, Y. (2022). Billion-user Customer Lifetime Value Prediction: An Industrial-scale Solution from Kuaishou. Proceedings of the 31st ACM International Conference on Information & Knowledge Management. https://doi.org/10.1145/3511808.3557152
Wang, X., Liu, T., & Miao, J. (2019). A Deep Probabilistic Model for Customer Lifetime Value Prediction. arXiv: Applications.
Su, H., Du, Z., Li, J., Zhu, L., & Lu, K. (2023). Cross-Domain Adaptative Learning for Online Advertisement Customer Lifetime Value Prediction. **, 4605-4613. https://doi.org/10.1609/aaai.v37i4.25583
Zhou, Z., Lin, L., Wang, H., Zhou, X., Wei, G., & Wang, S. (2024). A Cross Domain Method for Customer Lifetime Value Prediction in Supply Chain Platform. Proceedings of the ACM Web Conference 2024. https://doi.org/10.1145/3589334.3645391
Pan, Z., Lou, X., Jin, X., Ou, C., Liu, F., Zeng, T., He, C., Liu, X., Wei, L., & Wang, J. (2025). Progressive Tasks Guided Multi-Source Network for Customer Lifetime Value Prediction in Online Advertising. Proceedings of the Eighteenth ACM International Conference on Web Search and Data Mining. https://doi.org/10.1145/3701551.3703533
Wu, C., Jia, Q., Dong, Z., & Tang, R. (2023). Customer Lifetime Value Prediction: Towards the Paradigm Shift of Recommender System Objectives. Proceedings of the 17th ACM Conference on Recommender Systems. https://doi.org/10.1145/3604915.3609499
Liu, W., Xu, G., Ye, B., Luo, X., He, Y., & Yin, C. (2024). MDAN: Multi-distribution Adaptive Networks for LTV Prediction. **, 409-420. https://doi.org/10.1007/978-981-97-2259-4_31
Cao, X., Xu, Y., & Yang, X. (2024). Customer Lifetime Value Prediction with Uncertainty Estimation Using Monte Carlo Dropout. ArXiv, abs/2411.15944. https://doi.org/10.48550/arXiv.2411.15944
Ejgerdi, N., & Kazerooni, M. (2023). A stacked ensemble learning method for customer lifetime value prediction. Kybernetes, 53, 2342-2360. https://doi.org/10.1108/k-12-2022-1676
Kanchanapoom, K., & Chongwatpol, J. (2023). Integrated customer lifetime value models to support marketing decisions in the complementary and alternative medicine industry. Benchmarking: An International Journal. https://doi.org/10.1108/bij-07-2021-0431
Chamberlain, B., Cardoso, Â., Liu, C., Pagliari, R., & Deisenroth, M. (2017). Customer Lifetime Value Prediction Using Embeddings. Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. https://doi.org/10.1145/3097983.3098123
Zhang, W., Feng, J., & Li, F. (2024). Deep Learning-Based Customer Lifetime Value Prediction in Imbalanced Data Scenarios: A Case Study. **, 209-218. https://doi.org/10.1007/978-981-97-7184-4_18
Yang, X., Jia, B., Wang, S., & Zhang, S. (2023). Feature Missing-aware Routing-and-Fusion Network for Customer Lifetime Value Prediction in Advertising. Proceedings of the Sixteenth ACM International Conference on Web Search and Data Mining. https://doi.org/10.1145/3539597.3570460
Calabourdin, A., & Aksenov, K. (2023). Streaming Bayesian Modeling for predicting Fat-Tailed Customer Lifetime Value. ArXiv, abs/2312.00373. https://doi.org/10.48550/arXiv.2312.00373
Yan, Y., & Resnick, N. (2023). A high-performance turnkey system for customer lifetime value prediction in retail brands. Quantitative Marketing and Economics. https://doi.org/10.1007/s11129-023-09272-x
Xing, M., Bian, S., Zhao, W., Xiao, Z., Luo, X., Yin, C., Cai, J., & He, Y. (2021). Learning Reliable User Representations from Volatile and Sparse Data to Accurately Predict Customer Lifetime Value. Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining. https://doi.org/10.1145/3447548.3467079