Mobile Application Recommendation Based on Graph Embedding Depth Model

Nowadays, the information network has entered the era of big data, and more and more mobile application software has emerged. However, some applications do not meet the quality standards or do not match the description. The traditional recommendation work for mobile application software mainly includes coordinated filtering or content-based recommendation, but does not involve embedding graph structures. In response to such problems, this paper proposes a graph embedding deep prediction model PFore for mobile application rating recommendation. It uses the LINE model to extract the interaction relationship between users and applications, solve the problem of data sparsity, and use a factor decomposition machine to obtain first-order and second-order feature interactions. This article samples over 16000 applications in the Android market as data. The experiment shows that PFore improves MAE and RMSE by approximately $\mathbf{1 4. 7 \%}$ and $\mathbf{2 6. 1 \%}$ compared to the state of the art comparison model.