The paper introduces DeepFM, a neural network model for click-through rate (CTR) prediction, which combines the strengths of factorization machines (FM) and deep learning. DeepFM aims to learn both low- and high-order feature interactions in an end-to-end manner without requiring extensive feature engineering. The model consists of two components: the FM component for capturing pairwise feature interactions and the deep component for learning higher-order interactions. Unlike the Wide & Deep model, DeepFM shares the same input and embedding vector between its wide and deep parts, making it more efficient and effective. Extensive experiments on benchmark and commercial datasets demonstrate that DeepFM outperforms existing models in terms of CTR prediction accuracy and computational efficiency. The paper also discusses the relationship between DeepFM and other deep models, highlighting its unique advantages in handling both low- and high-order feature interactions.The paper introduces DeepFM, a neural network model for click-through rate (CTR) prediction, which combines the strengths of factorization machines (FM) and deep learning. DeepFM aims to learn both low- and high-order feature interactions in an end-to-end manner without requiring extensive feature engineering. The model consists of two components: the FM component for capturing pairwise feature interactions and the deep component for learning higher-order interactions. Unlike the Wide & Deep model, DeepFM shares the same input and embedding vector between its wide and deep parts, making it more efficient and effective. Extensive experiments on benchmark and commercial datasets demonstrate that DeepFM outperforms existing models in terms of CTR prediction accuracy and computational efficiency. The paper also discusses the relationship between DeepFM and other deep models, highlighting its unique advantages in handling both low- and high-order feature interactions.