This paper proposes a novel deep matrix factorization model for top-N recommendation, combining explicit ratings and implicit feedback. The model constructs a user-item matrix with both explicit ratings and non-preference implicit feedback, then uses a deep neural network architecture to learn a common low-dimensional space for users and items. A new loss function based on binary cross entropy is designed to incorporate both explicit ratings and implicit feedback for better optimization. The experimental results show that the proposed model outperforms other state-of-the-art methods on several benchmark datasets. The model's effectiveness is demonstrated through extensive experiments, including different experimental settings. The model's architecture is inspired by deep structured semantic models, which have been proven useful for web search. The model uses a neural network to map users and items into a latent space, and a new loss function is designed to consider both explicit and implicit feedback. The model's performance is evaluated using metrics such as NDCG and HR, and it achieves significant improvements over existing methods. The model's input matrix includes both explicit ratings and implicit feedback, and the experiments show that the input matrix is crucial for the model's performance. The model is also sensitive to hyper-parameters such as the negative sampling ratio and the number of layers in the neural network. The results show that the model performs best with a negative sampling ratio of around 5 and two layers in the neural network. The model's final latent space is also sensitive to the number of factors, with the best performance achieved when the number of factors is around 64. The model's effectiveness is demonstrated through extensive experiments, and it is shown to be a promising approach for top-N recommendation. The paper concludes that the proposed model is a novel approach for recommendation systems, combining explicit ratings and implicit feedback, and that future work could include extending the model with pairwise objective functions and incorporating auxiliary data.This paper proposes a novel deep matrix factorization model for top-N recommendation, combining explicit ratings and implicit feedback. The model constructs a user-item matrix with both explicit ratings and non-preference implicit feedback, then uses a deep neural network architecture to learn a common low-dimensional space for users and items. A new loss function based on binary cross entropy is designed to incorporate both explicit ratings and implicit feedback for better optimization. The experimental results show that the proposed model outperforms other state-of-the-art methods on several benchmark datasets. The model's effectiveness is demonstrated through extensive experiments, including different experimental settings. The model's architecture is inspired by deep structured semantic models, which have been proven useful for web search. The model uses a neural network to map users and items into a latent space, and a new loss function is designed to consider both explicit and implicit feedback. The model's performance is evaluated using metrics such as NDCG and HR, and it achieves significant improvements over existing methods. The model's input matrix includes both explicit ratings and implicit feedback, and the experiments show that the input matrix is crucial for the model's performance. The model is also sensitive to hyper-parameters such as the negative sampling ratio and the number of layers in the neural network. The results show that the model performs best with a negative sampling ratio of around 5 and two layers in the neural network. The model's final latent space is also sensitive to the number of factors, with the best performance achieved when the number of factors is around 64. The model's effectiveness is demonstrated through extensive experiments, and it is shown to be a promising approach for top-N recommendation. The paper concludes that the proposed model is a novel approach for recommendation systems, combining explicit ratings and implicit feedback, and that future work could include extending the model with pairwise objective functions and incorporating auxiliary data.