CDF-LS: Contrastive Network for Emphasizing Feature Differences with Fusing Long- and Short-Term Interest Features

Modelling both long- and short-term user interests from historical data is crucial for generating accurate recommendations. However, unifying these metrics across multiple application domains can be challenging, existing approaches often rely on complex, intertwined models which difficult to interpret. To address this issue, we propose a lightweight, plug-and-play interest enhancement module that fuses vectors two independent models. After analyzing the dataset, identify deviations in recommendation performance of compensate differences, use feature loss correction during training. In fusion process, explicitly split long-term features with longer duration into local features. We then shared attention mechanism fuse obtain interaction correct bias between models, introduce comparison learning task monitors similarity features, This adaptively reduces distance similar Our proposed combines compares domain datasets. As result, it not only accelerates convergence but also achieves outstanding challenging scenarios.