Philip kotler, a master of management, thinks: "The user's behavior trajectory mainly includes five stages: demand generation, information collection, scheme comparison, purchase decision and post-purchase behavior." Following this rule, the data of readers' behavior is mainly generated by five steps: reading service demand, information collection related to reading activities, selection of reading content and service mode, reading activities and user reading feedback. Libraries can collect, process, calculate, analyze and make decisions on readers' behavior big data information according to the life cycle law of readers' reading activities, and provide big data decision support for users' service mode selection and process.
The library's collection of data on readers' demand for reading services mainly involves research data on library readers' demand, reading demand information fed back by readers, readers' browsing footprints on the website, readers' comments, retrieval history, borrowing history, readers' selection and deletion of service content, readers' subscription and other behavioral data. The collection of information related to reading activities includes readers' individual characteristics, reading habits, reading terminal types, working modes and other data. The choice of reading content and service mode mainly involves data such as library user service mode, user service mode and content, reading application type and working mode, reading activity mode selection and changing trend. Reading activities are mainly composed of website visit logs, searching and downloading of reading contents, recording of readers' reading behaviors by the server, reading frequency, total online time of readers, searching and browsing of reading contents, classification of reading contents, reading social relations and friend interaction, location information of mobile reading terminals, reading-related behaviors on third-party websites and other data. Reading feedback behavior is mainly composed of data such as reader's reading experience, user satisfaction evaluation, reader's loyalty, and reader's message evaluation.
2. Important issues that should be paid attention to in the analysis of readers' behavior.
(1) The big data noise in the process of reader behavior analysis interferes with its effectiveness.
Readers' behavior data should be accurate, effective and reliable, and the library should expand the breadth and depth of readers' behavior data collection as much as possible. However, with the increase of the scope and depth of data collection, it is bound to bring a lot of invalid behavior data, which greatly affects the reliability and availability of reader behavior data and increases the complexity and cost of analyzing reader behavior data. Therefore, it is necessary to clean up the collected data and filter out the noise data that interferes with readers' behavior analysis to ensure the safety and reliability of the data. Noise data generally comes from three aspects: first, the garbage data generated by readers' misoperation; Second, different data analysis methods, angles and tools of the same set of data affect the accuracy and efficiency of the data; Thirdly, the performance of software and hardware systems and the quality of operators have also become the real factors of data interference.
(2) The analysis of readers' behavior data provides support for readers' personalized service.
Ensuring readers' high user experience satisfaction and service fairness is the goal of personalized service for library readers. On the basis of analyzing readers' individual behaviors, libraries should complete personalized recommendations on readers' preferences, reading methods, reading contents and reading terminals, so as to ensure the reasonable allocation of resources according to readers' identities and reading needs, and realize the optimal allocation of resources and the fairness of services. At the same time, we should also promote reading content and information services in a targeted manner to meet the individual needs of readers and reduce the time for readers to search and acquire knowledge. In addition, through the big data analysis of users' behaviors, the credibility and safety of readers' behaviors can be guaranteed, malicious behaviors of illegal users can be found and stopped in time, the utilization rate of library resources can be maximized, and the safety and reliability of the service platform can be guaranteed.
The premise of personalized service is to extract users' sexual attributes and analyze the data. Before data analysis, we need to find the range of user preference dimensions. By tracking the user's behavior, we can grasp all dimensions of the user himself and the user's target behavior. Defining dimensions is the basis of accumulating data. All personalized presentations are the comprehensive results of users' multi-dimensional choices, and the diversity of results becomes personalization. The criterion for selecting dimensions is that users can easily branch and select dimensions, while non-decision or non-branch data can be stored separately as secondary data, and finally the detailed behavior decision path is retained, and dimensions are extracted according to this path.
(3) Technical challenges in the process of reader behavior analysis.
First of all, the reader behavior data collected by the library in the big data environment has the characteristics of mass, multi-type, rapid increase and real-time processing. Therefore, the reader behavior analysis puts forward higher requirements for the massive data storage, management and real-time high-speed processing capabilities of the big data platform. Secondly, readers' behavior data has the characteristics of unstructured, vague and disordered, and it is impossible to express the data structure and meaning in a standardized way. Thirdly, the mining, analysis and decision-making of readers' behavior data is a long-term and gradual accumulation process, which requires big data platform to ensure that the analysis system has strong scalability and dynamic updating ability in the process of behavior data storage, resource organization, information integration and knowledge discovery according to the life cycle law of readers' behavior analysis process. Finally, the process of reader behavior analysis puts forward higher requirements for users' reading privacy protection, data security management and usability guarantee, decision-making and service system security, and reader's reading experience QOS (quality of service) guarantee.
(4) Guide the whole process of readers' behavior analysis with readers' personalized service demand.
First of all, the library should monitor and collect data from server logs, monitoring system, comment feedback system and online customer service system, including massive storage of readers' reading activities, website access paths, access and reading contents, service resource attention, reading comments and satisfaction feedback. And refine the collected user behavior data to describe the behavior of readers. Secondly, the library should scientifically judge readers' personalized service demand and the best service mode through real-time calculation and comprehensive analysis of reading terminal type, reading terminal geographical location, user location movement data, user feature data, personalized service history data and user demand feedback data, and meet users' personalized reading demand on the premise of ensuring the optimal allocation of service resources and the highest service efficiency; Thirdly, through the real-time analysis of readers' behavior data, we can grasp the changing trend of readers' personalized reading needs, and dynamically adjust and improve readers' service strategies and guarantee modes. In addition, it is also an important content and goal of library reader behavior analysis to accurately judge the dangerous behavior in abnormal readers' reading activities, improve the library's security prevention ability and reduce risks.