ISSN: 2222-6990
Open access
Click-through rate (CTR) prediction is central to marketing decision-making, where predicted probabilities are directly used for bidding, filtering, and budget allocation. Practical systems therefore require not only strong ranking performance, but also reliable probabilities that remain stable under noisy fields, shifting contexts, and redundant fused representations. However, most existing CTR models primarily optimize ranking accuracy, while paying less attention to probability reliability and stability, which limits their usefulness in real-world marketing decision-making. In this work, we propose SwaDeepFM, a stage-wise attention CTR model based on a DeepFM backbone. SwaDeepFM introduces three lightweight modules at different stages of the prediction pipeline. First, SE performs field-wise reweighting at the embedding stage to suppress noisy fields before interactions are formed. Second, a Context Transformer (CoT) aggregates context-conditioned dependencies among field tokens to stabilize high-order interaction selection. Third, CBAM refines the fused representation before prediction to remove redundancy and improve probabilistic robustness. Experiments on three public CTR benchmarks, Criteo, Avazu, and KDD12, against several representative baselines show consistent improvements in AUC and LogLoss. Fixed ablations further confirm that each module contributes complementary gains. Additional analyses based on field-weight visualization, attention inspection, long-tail bucketing, and calibration and stability protocols explain when the method helps most and why the resulting probabilities are more reliable for marketing decisions.
Aden, & Wang, Y. (2012). KDD Cup 2012, track 2. Kaggle. ?https://kaggle.com/competitions/kddcup2012-track2?
Guan, F., Zhan, J., & Yang, J. (2025). Accurate and interpretable CTR prediction via distilled neural additive feature interaction network. Journal of Big Data, 12.
Guo, H., Tang, R., Ye, Y., Li, Z., He, X., & Dong, Z. (2018). DeepFM: An end-to-end wide & deep learning framework for CTR prediction. arXiv preprint arXiv:1804.04950.
Dai, Q., Xiao, J., Du, Z., Zhu, J., Luo, C., Wu, X.-M., & Dong, Z. (2025). MCNet: Monotonic calibration networks for expressive uncertainty calibration in online advertising. Proceedings of the ACM on Web Conference 2025.
Qu, Y., Fang, B., Zhang, W., Tang, R., Niu, M., Guo, H., Yu, Y., & He, X. (2018). Product-based neural networks for user response prediction over multi-field categorical data. ACM Transactions on Information Systems (TOIS), 37, 1–35.
Škrlj, B., Karni, Y., Gaspersic, G., Mramor, B., Stolin, Y., Jakomin, M., ... & Klein, A. (2025). DCN^2: Interplay of implicit collision weights and explicit cross layers for large-scale recommendation. arXiv preprint arXiv:2506.21624.
Tien, J.-B., joycenv, & Chapelle, O. (2014). Display advertising challenge. Kaggle. ?https://kaggle.com/competitions/criteo-display-ad-challenge?
Wang, S., & Cukierski, W. (2014). Click-through rate prediction. Kaggle. ?https://kaggle.com/competitions/avazu-ctr-prediction?
Wang, Y., Ji, H., He, X., Yu, J., Han, H., Zhai, R., & Wang, L. (2024). Disentangled self-attention neural network based on information sharing for click-through rate prediction. PeerJ Computer Science, 10.
Wu, S., Du, L., Yang, J.-Q., Wang, Y., Zhan, D.-C., Zhao, S., & Sun, Z. (2023). RE-SORT: Removing spurious correlation in multilevel interaction for CTR prediction (pp. 3816–3828).
Woo, S., Park, J., Lee, J.-Y., & Kweon, I. S. (2018). CBAM: Convolutional block attention module. Proceedings of the European Conference on Computer Vision (ECCV), 3–19.
Zhang, Y., Cheng, X., Wei, W., & Meng, Y. (2025). Deep double towers click through rate prediction model with multi-head bilinear fusion. Symmetry, 17, 159.
Zhang, Y., Shi, T., Feng, F., Wang, W., Wang, D., He, X., & Zhang, Y. (2023). Reformulating CTR prediction: Learning invariant feature interactions for recommendation. Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval.
Bao, Q., & Farhana, N. (2026). SwaDeepFM: Reliable CTR Prediction for Marketing Decision-Making via Stage-wise Attention Selection and Post-fusion Redundancy Removal. International Journal of Academic Research in Business and Social Sciences, 16(4), 1030–1041.
Copyright: © 2026 The Author(s)
Published by HRMARS (www.hrmars.com)
This article is published under the Creative Commons Attribution (CC BY 4.0) license. Anyone may reproduce, distribute, translate and create derivative works of this article (for both commercial and non-commercial purposes), subject to full attribution to the original publication and authors. The full terms of this license may be seen at: http://creativecommons.org/licences/by/4.0/legalcode