Se-xDeepFEFM: Combining Low-Order Feature Refinement and Interaction Intensity Evaluation for Click-Through Rate Prediction

Click-through rate (CTR) prediction can provide considerable economic and social benefits. Few studies have considered the importance of low-order features, usually employing a simple feature interaction method. To address these issues, we propose novel model called Senet extreme deep field-embedded factorization machine (Se-xDFEFM) for more effective CTR prediction. We first embed squeeze-excitation network (Senet) module into Se-xDFEFM to complete refinement, which better filter noisy information. Then, implement our (FEFM) learn symmetric matrix embeddings each field pair, along with single-vector feature, builds firm foundation subsequent interaction. Finally, design compressed (CIN) realize construction definite order through vector-wise use neural (DNN) CIN simultaneously but complementary explicit implicit interactions. Experimental results demonstrate that outperforms other state-of-the-art baselines. Our is robust Importantly, variants also achieve competitive recommendation performance, demonstrating their scalability.