Visual Analytics Of Employee Behaviors Based On Enterprise Log Metadata

Enterprise log metadata,which is the record saved by the system when employees use the network service within the enterprise,including web browsing log metadata,mail log metadata,and TCP traffic log metadata.To a certain extent,these metadata reflect the organizational structure,community groups,daily work patterns,and various abnormal behaviors of employees within the company.The analysis of these metadata helps the company's senior management to control the operation of the enterprise in a timely manner,identify potential threats to the enterprise,and make better decisions.Most of the existing analysis methods use algorithms such as data mining and machine learning to analyze on a single data basis such as mail logs.This paper believes that:(1)combining data-centric analysis algorithms with human-centered interactive visual analysis methods can simultaneously exploit the advantages of algorithms and humans;(2)visual analysis methods can be more effective The log metadata of massive,multivariate,time-varying,heterogeneous and other characteristics are combined and analyzed to provide multi-angle and all-round deep analysis.The main research work of this paper is to present the data to knowledge transformation process through a series of data processing and visual mapping,and to better promote data mining by means of interactive visual analysis.Specifically,the enterprise log metadata is used as the information source,and the theoretical knowledge of visualization and visual analytics is applied to the analysis of the problem of enterprise log metadata.To this end,this paper usesthe visualization and visual analysis technology to design and implement the EWB-VIS,a visual analysis system for employee work behavior for enterprise log metadata.It mainly provides three visual analysis methods:(1)A method for distinguishing employee work department visualization based on clustering algorithm.The method combines the force-oriented layout algorithm,and uses the scatter plot method to flexibly and intuitively display the employees and their departments;(2)the Gantt chart visualization method that fuses the behavior information time axis.The method can accurately display the data with time information,and display as much data as possible to the user in combination with the stacked graph to realize the analysis of the time series event pattern;(3)the auxiliary visualization method.It provides users with multi-level optional visualization methods,including displaying the text data of employee web browsing and sending and receiving emails with word cloud diagram,using radar chart to display the activities of employees in the time dimension and the number of activities,and displaying the internals of the company with thermal maps.TCP traffic for network activity.In addition,a rich interaction is designed for the system to achieve a linkage analysis of employee work behavior related information.Finally,the effectiveness of the system and the effectiveness of the related visualization methods are demonstrated through experiments on the open dataset of the ChinaVis2018 Challenge.The visualization system of this paper is mainly for analysts who have knowledge of certain enterprise log metadata,and can intuitively support them to analyze employee work behavior patterns and discover employee anomalies.Compared with pure data mining algorithms,this paper combines data-centric analysis algorithms with human-centered interactive visualization to make use of novel view design and rich interactions while leveraging the advantages of algorithms and human analysis.The analysis process is more intuitive and easy to understand,and better solves many problems in the analysis of employees' work behavior patterns.