ROI-Optimized Sequential Advertising Recommendation Framework with Cloud-Native Scalable Data Infrastructure

Bingqing Ni; Hanwu Li; Jingwei Zhang

doi:10.63313/JCSFT.9068

Authors

Bingqing Ni Chongqing University, Chongqing, China Author
Hanwu Li Amazon.com Services LLC, 98004, USA Author
Jingwei Zhang Stanford University, Stanford, CA, 94305, USA Author

DOI:

https://doi.org/10.63313/JCSFT.9068

Keywords:

Advertising Recommendation, ROI Optimization, Sequential Modeling, Multi-task Learning, Cloud-Native Systems, Real-time Recommendation

Abstract

Online advertising recommendation systems are fundamental to large-scale e-commerce platforms, where accurate user targeting and efficient system deployment directly impact revenue generation and user experience. Existing approaches primarily optimize click-through rate (CTR) or conversion rate (CVR), which often fail to align with business-oriented objectives such as return on investment (ROI), while also lacking scalability under cloud-native, real-time environments. To address these challenges, we propose ROSA-Rec, a ROI-optimized sequential advertising recommendation framework that tightly integrates user behavior modeling with a cloud-native scalable data infrastructure. Specifically, ROSA-Rec employs a behavior-aware Transformer-based encoder to capture dynamic user interests from heterogeneous behavioral sequences, incorporating both action importance and temporal decay. A cross-attention interaction module is further designed to enhance fine-grained matching between users and advertisements. To directly optimize business value, we introduce a ROI-aware multi-task learning objective that jointly models CTR, CVR, and revenue signals. In addition, a hybrid online-offline serving architecture built on cloud-native technologies enables real-time feature updates and low-latency inference, ensuring scalability and consistency in large-scale production settings.Extensive experiments on benchmark and semi-simulated advertising datasets demonstrate that ROSA-Rec achieves superior performance, including approximately a 10.5% improvement in ROI over the strongest baseline (SASRec), while also improving CTR AUC from 0.826 to 0.854 and CVR-AUC from 0.748 to 0.781. These results confirm that ROSA-Rec effectively enhances both predictive accuracy and business-oriented metrics under large-scale advertising scenarios.

References

[1] Rendle S, Freudenthaler C, Schmidt-Thieme L. Factorizing personalized markov chains for next-basket recommendation[C]//Proceedings of the 19th international conference on World wide web. 2010: 811-820.

[2] Hidasi B, Karatzoglou A, Baltrunas L, et al. Session-based recommendations with recurrent neural networks[J]. arXiv preprint arXiv:1511.06939, 2015.

[3] Kang W C, McAuley J. Self-attentive sequential recommendation[C]//2018 IEEE international conference on data mining (ICDM). IEEE, 2018: 197-206.

[4] Sun F, Liu J, Wu J, et al. BERT4Rec: Sequential recommendation with bidirectional encoder representations from transformer[C]//Proceedings of the 28th ACM international conference on information and knowledge management. 2019: 1441-1450.

[5] Zhou G, Zhu X, Song C, et al. Deep interest network for click-through rate prediction[C]//Proceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining. 2018: 1059-1068.

[6] Richardson M, Dominowska E, Ragno R. Predicting clicks: estimating the click-through rate for new ads[C]//Proceedings of the 16th international conference on World Wide Web. 2007: 521-530.

[7] Ma X, Zhao L, Huang G, et al. Entire space multi-task model: An effective approach for estimating post-click conversion rate[C]//The 41st International ACM SIGIR Conference on Research & Development in Information Retrieval. 2018: 1137-1140.

[8] Zhao J, Qiu G, Guan Z, et al. Deep reinforcement learning for sponsored search real-time bidding[C]//Proceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining. 2018: 1021-1030.

[9] Li X, Li Z, Lin X. Automated Implementation of Machine Learning-Based Causal Inference in Product Operations Decision Making[C]//Proceedings of the 2nd International Symposium on Integrated Circuit Design and Integrated Systems. 2025: 204-209.

[10] Abadi M, Barham P, Chen J, et al. {TensorFlow}: a system for {Large-Scale} machine learning[C]//12th USENIX symposium on operating systems design and implementation (OSDI 16). 2016: 265-283.

[11] Carbone P, Katsifodimos A, Ewen S, et al. Apache flink: Stream and batch processing in a single engine[J]. The Bulletin of the Technical Committee on Data Engineering, 2015, 38(4).

[12] MJ J K. Feature Store for Machine Learning: Curate, discover, share and serve ML features at scale[M]. Packt Publishing Ltd, 2022.

[13] Wilder B, Ewing E, Dilkina B, et al. End to end learning and optimization on graphs[J]. Advances in Neural Information Processing Systems, 2019, 32.

[14] Wang L, Zhang X, Jiang L. Prediction Framework for E-Commerce Platform Sales Data Based on Informer: A Study on Furniture Sales on Amazon E-Commerce Platform[J]. Economics and Management Innovation, 2025, 2(6): 80-87.

[15] Zhang Y, Bai Z. GenRiskNet: A GenAI-Driven Multi-Source Heterogeneous Data Fusion Framework for Financial Risk Prediction[J]. Economics and Management Innovation, 2026, 3(1): 112-121.

[16] Zhang S, Yang S, Zhang W, et al. Hybrid Beamforming for Subarray-Level Movable Antenna Enhanced MU-MIMO Communications[J]. IEEE Wireless Communications Letters, 2026, 15: 2559-2563.

[17] Li Z, Li X, Lin X. Design and Implementation of a Platform for Business Intelligence Knowledge Mining and Graph Construction Based on Deep Learning[C]//Proceedings of the 2nd International Symposium on Integrated Circuit Design and Integrated Systems. 2025: 137-141.

[18] Liang Z, Wei W, Zhang K, et al. Research on multi-hop inference optimization of llm based on mquake framework[J]. arXiv preprint arXiv:2509.04770, 2025.