With the advent of the era of big data, on the one hand, the traditional, single and rough information service model has been difficult to meet the needs of readers' personalized information service; On the other hand, with the rapid growth of the number of readers and the data volume of readers' behaviors, the traditional data storage architecture and management mode have been seriously lagging behind, and it is impossible to complete multi-dimensional and all-round data collection, storage, organization and analysis, and it is impossible to screen and filter out tendentious information that can accurately predict readers' interests and personalized information service needs from a large number of complex data. Due to the lack of effective data support, the library can not complete the most basic and effective personalized information service push for readers. Therefore, the library's intelligent service platform should have several functions: First, it should ensure the safe, efficient and accurate management of data collection, transmission, processing and analysis, standardize the collected reader behavior data, ensure the systematicness, integrity and value of the data, and provide a solid and reliable guarantee for accurately predicting readers' future reading tendency and providing personalized information services. Second, make a multi-dimensional analysis of readers, establish analysis dimensions based on readers' specialties, reading habits, reading methods and other related attributes, sort out, screen and classify all literature information resources, and establish an operable relational database with readers' multi-dimensional data to ensure that the personalized information push received by each reader is related to its attributes. Third, establish a reader behavior analysis system to comprehensively analyze data such as reader composition, reading habits, reading methods, content preferences, and reader credit. To explore the relationship between readers, manage and analyze it accordingly, and establish a reader behavior analysis system of "reader-resource-channel".