PTUM: Pre-training User Model from Unlabeled User Behaviors via Self-supervision
Chuhan Wu, Fangzhao Wu, Tao Qi, Jianxun Lian, Yongfeng Huang, Xing Xie
TL;DR
The paper introduces PTUM, a self-supervised pre-training framework that learns universal user representations from unlabeled behavior data using two tasks: Masked Behavior Prediction and Next K Behaviors Prediction. By pre-training on large-scale unlabeled histories and then fine-tuning on downstream tasks like demographic prediction and CTR, PTUM consistently improves performance, especially when labeled data are scarce. Ablation studies show both MBP and NBP contribute meaningfully, with the best results achieved when both tasks are combined. The work demonstrates the viability and practical impact of self-supervised pre-training for user modeling in personalized web services.
Abstract
User modeling is critical for many personalized web services. Many existing methods model users based on their behaviors and the labeled data of target tasks. However, these methods cannot exploit useful information in unlabeled user behavior data, and their performance may be not optimal when labeled data is scarce. Motivated by pre-trained language models which are pre-trained on large-scale unlabeled corpus to empower many downstream tasks, in this paper we propose to pre-train user models from large-scale unlabeled user behaviors data. We propose two self-supervision tasks for user model pre-training. The first one is masked behavior prediction, which can model the relatedness between historical behaviors. The second one is next $K$ behavior prediction, which can model the relatedness between past and future behaviors. The pre-trained user models are finetuned in downstream tasks to learn task-specific user representations. Experimental results on two real-world datasets validate the effectiveness of our proposed user model pre-training method.