FAN: Fatigue-Aware Network for Click-Through Rate Prediction in E-commerce Recommendation

Since clicks usually contain heavy noise, increasing research efforts have been devoted to modeling implicit negative user behaviors (i.e., non-clicks). However, they either rely on explicit (e.g., dislikes) or simply treat non-clicks as feedback, failing learn interests comprehensively. In such situations, users may experience fatigue because of seeing too many similar recommendations. this paper, we propose Fatigue-Aware Network (FAN), a novel CTR model that directly perceives from non-clicks. Specifically, first apply Fourier Transformation the time series generated non-clicks, obtaining its frequency spectrum which contains comprehensive information about fatigue. Then is modulated by category target item bias both upper bound and users’ patience different for categories. Moreover, gating network adopted confidence an auxiliary task designed guide learning fatigue, so can obtain well-learned representation combine it with final prediction. Experimental results real-world datasets validate superiority FAN online A/B tests also show outperforms representative models significantly.