Predicting User Behaviors with Scene via Dual Sequence Networks
Keywords: dual sequence networks, scene-aware, behavior prediction
Abstract: Modeling sequential user behaviors for future action prediction is crucial in improving user's information retrieval experience. Recent studies highlight the importance of incorporating contextual information to enhance prediction performance. One crucial and typical contextual information is the scene feature which we define it as sub-interfaces within an app, created by designers to provide specific functionalities, such as ''text2product search" and ''live" in e-commence apps. Different scenes exhibit distinct functionalities and usage habits, leading to significant distribution gap in user engagement across them. Popular sequential behavior models either ignore the scene feature or merely use it as attribute embeddings, which could lead to substantial information loss or cannot capture the interplay between scene and item in modeling dynamic user interests. In this work, we propose a novel Dual Sequence Prediction network (DSPnet) to effectively capture the interplay between scene and item sequences for future behavior prediction. DSPnet consists of two parallel networks dedicated to predicting scene and item sequences, and a sequence feature enhancement module to capture the interplay. Further, considering the randomness and noise in learning sequence dynamics, we introduce Conditional Contrastive Regularization (CCR) loss to capture the invariance of similar historical sequences. Theoretical analysis suggests that DSPnet can learn the joint relationships between scene and item sequences, and also show better robustness on real-world user behaviors. Extensive experiments are conducted on one public benchmark and two collected industrial datasets. The codes and collected datasets will be made public soon.
Primary Area: applications to computer vision, audio, language, and other modalities
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 4288
Loading