Design And Implementation Of Spatiotemporal Operators Of Analysis For Traffic Management Data

The traffic management big data is an important part of space-time big data.Since the reform and opening up,China's urban transportation has developed rapidly.And we have accumulated massive traffic data,including static data and dynamic data.These traffic data contains a lot of valuable information,which helps the traffic control department to make decisions,and accurately assist the traffic management work.In the traffic daily work,it is often required to count the illegal rankings of the drivers,to count the number of accidents in the road sections and to analyze the track point and vehicle information of the overspeed on road network.The essence of these work is query statistics about related data tables in database.SQLstatements are simple or complex.However,there are many similar jobs in the management of traffic.If every statistical analysis task is done by the database query statistics or the corresponding program,the workload of the traffic management is huge and the efficiency is obviously low.Therefore,on the basis of the analysis of the application value of the traffic big data and the demand for management,it is an urgent problem to calculate and analyze the information efficiently in the traffic big data.However the traditional space-time data analysis system is mostly running on the client,lack of online service,low support for spatial data type and low calculate and analysis of mass data.Based on the requirements and existing problems of traffic work,a space-time operator for large data analysis is attempted in this paper,which aims to decompose the traffic data analysis into multiple sub tasks.In this paper,space-time operator is related to sub tasks to express the data operation of sub tasks,and the analysis of traffic data is ed into a model,so that the whole model construction is composed of multiple sub tasks with corresponding space-time operator.Each operator is responsible for the function of the sub task,and there are dependencies in sub tasks.The execution of the whole model is the execution of sub tasks according to the dependent relationship.Based on the above ideas,we design related space-time operators.According to the characteristics and needs of daily traffic work,the types of space-time operators are divided in detail.In the aspact of operator execution,this paper carries out the corresponding execution efficiency experiments on the Spark execution strategy and the SQL structured record execution strategy.Based on the comparision experiment results and the operator execution efficiency,the space-time operator execution strategy is combined with two execution strategies,and the selection rules of operator execution strategy are designed.In the execution process,we choose the appropriate execution strategy to calculate.Considering the two execution strategies,the concept of virtual table is defined as input and output parameter of space-time operator,and the virtual table structure is designed according to the characteristics of two execution strategies.For the computation of operators,we design basic parameters and the parameters of space-time operators,which refer to the basic parameters.In addition,we define and design the unified computing interface of operators and the computing interfaces under different execution strategies.The model is composed of multiple sub tasks.The execution of models is the schedules and execution of multiple sub tasks.In the aspect of model building,this paper constructs the model by combining the Dexecutor scheduling framework with the model and asocciating tasks with space-time operators.Also this paper expounds the construction process of the model execution plan.In the terms of model execution,this paper descibes the whole execution and single step execution of model,and describes the scheduling execution process of multiple sub tasks in a model.In order to elaborate the scheduling process,this paper analyze the scheduling process from two aspects of non space and space by example of typical model.Finally,the efficiency and advantages of model scheduling execution are verified by comparing the execution efficiency of the scheduling and serial execution of multiple sub tasks in a model.In addition,based on the above research results,the prototype system has been designed.Also this paper designs the corresponding experiments and the experimental analysis is carried out on the above research,and the expected effect is achieved.Finally,the work of this article is descibed as a whole,the research content of this article is summarized,the shortcomings of this paper are pointed out,and the follow-up research focus is put forward to the insuffient places.