Research on Deep Clustering and Multimodal Deep Learning Service Recommendation System for Large-Scale User Data

With the advancement of Internet technology and the growth of user data, accurately predicting user needs and providing personalized recommendations have become key challenges for intelligent recommendation systems. The diversity and sparsity of large-scale user data increase the complexity of demand prediction. To this end, this paper proposes a service recommendation method that combines deep clustering and multimodal deep learning to improve recommendation accuracy and personalization capabilities. This study uses an unsupervised learning framework to reduce the dimension and normalize user data, reduce sparsity and optimize clustering results, and provide accurate user classification for subsequent prediction models. This paper proposes a demand prediction method based on soft attention mechanism and multimodal deep learning (SAMML). By fusing text and image features and combining soft attention mechanism to strengthen the focus on key information, accurate prediction is finally achieved through the gated recurrent unit (GRU) network. Experimental results show that the SAMML method improves the evaluation indicators such as MAE, MSE and R2 by 5.81%, 6.45% and 6.0% respectively, verifying its effectiveness in practical applications. This study provides new ideas and technical support for personalized recommendation of large-scale user data.