Recently, network technology and internationalization of scientific research have pushed researchers to turn to electronic literature resources. In the face of changes in user behavior, e-documentation platforms need to efficiently manage and deeply understand the sequence of user behavior in order to improve the accuracy and satisfaction of their services. Our study introduces an innovative methodology for analyzing user behavior within literature resource platforms through modeling sequential patterns of user actions. By leveraging the word2vec technique from natural language processing, we transform user behavior sequences comprising both the nature of the actions and the characteristics of the accessed documents into meaningful vector representations. This transformation enables the detection of anomalies in the behavioral patterns by examining clusters in the vector space. Based on this approach, we design and implement a visual analysis system of abnormal access: UBAViz. The system allows analysts to check and analyze the results of modeling user behavioral sequences and to improve the understanding of anomalous access users. Through two user cases, we demonstrate how our approach and system help to detect and analyze abnormal users, showing that the system allows managers to fully and carefully mine the behavioral patterns of different users.