Session-aware Item-combination Recommendation with Transformer Network
This is an incremental improvement for recommendation systems in e-commerce or content platforms, focusing on session-aware predictions.
The paper tackled the problem of item combination prediction in recommendation systems by developing a transformer-based network with session-aware reweighted loss and data augmentation techniques, achieving a categorization accuracy of 0.39224 and ranking 2nd in the IEEE BigData Cup 2021 competition.
In this paper, we detailedly describe our solution for the IEEE BigData Cup 2021: RL-based RecSys (Track 1: Item Combination Prediction). We first conduct an exploratory data analysis on the dataset and then utilize the findings to design our framework. Specifically, we use a two-headed transformer-based network to predict user feedback and unlocked sessions, along with the proposed session-aware reweighted loss, multi-tasking with click behavior prediction, and randomness-in-session augmentation. In the final private leaderboard on Kaggle, our method ranked 2nd with a categorization accuracy of 0.39224.