Multidimensional Analysis Of Daily Behaviors Data Based On Graph Model

With the rapid development and popularization of the Internet and computer technology,the 21st century has been ushered into the data-driven life.The digital traces are produced incessantly by the electronic equipment in the daily lives of the people.The activities recorded by digital traces are called the individual daily living behavior data,known as behavior data.The behavior data is not only closely related to people themselves,but also has a growing social,economical and practical value.In the meantime,the accumulation of behavior data is growing huge and complex,which generates the needs to analyze data from multiple views and multiple levels.However,there is currently no suitable data model to implement the multidimensional analysis on behavior data.Therefore,this thesis proposed a multidimensional framework to analyze behavior data from multiple perspectives,aiming at helping individuals understand themselves better by the behaviors in their daily lives and providing support for behavioral decisions.Firstly,based on the content characteristics of personal daily behavior data,the concept of heterogeneous sequence network was proposed in this thesis.The construction method of the network was also designed and implemented,and was applied to the scenario of behavior data analysis to construct behavior network.Secondly,in order to implement multidimensional analysis of heterogeneous sequence networks,a heterogeneous sequence graph cube model was put forward,known as HS Cube.Based on the heterogeneous sequence network,a novel dimension concept of structure dimensions was defined,which is a set of topological structures obtained by mapping the logical relation between data elements.The structure dimension included three levels:vertex dimensions,edge dimensions,and subgraph dimensions.The concept of measures was also redefined,which contains content measures,numeric measures and graph measures.Based on the structure dimension and measures,the calculation methods of aggregate graphs were described,and the definition of the dimension hierarchy was also introduced.Then all the aggregate graphs were constructed into a graph cube according to the dimension hierarchy.Thirdly,the semantics of the basic OLAP operations,that is Roll-up,Drill-down,and Slice/dice,were redefined for the HS Cube.At the same time,in order to provide the users with a friendly query interface,we formulated four cube query modes,the simple query,set query,sequence query and complex query respectively.Finally,the system architecture of HS Cube and the partial materialization strategy of graph cube were designed.We also introduced the realization algorithms of aggregate and OLAP operations.The graph cube model was implemented by the graph database Neo4j?and showed the model could provide the multidimensional analysis for behaviors data effectively.