Big Data Utilizing Strategy Of Intelligent Transport System In Mobile Cloud Platform

In recent years, the complexity of the transportation monitoring data is growing fast every day. The urban transportation problem is becoming more serious despite of the effort of road network evaluation and data analysis. One of the key points is how to make high utilization of large amount transportation data, parsing, mining and invoking the date in real time, providing helpful transportation serving information for the drivers.The thesis is based on the major demonstration project of Jiangsu Provincial Science and Technology Department- the demonstration system based on the detection, mining, convergence and publishing of the freeway traffic sensor network information and assisting decision. It constructs the transportation data cluster to analysis mass transportation data using the spark framework. It uses the improved genetic algorithm to do load balance and the server uses reactor model to supply high concurrency service. It calculates the status of the road network based on the GPS data which gathered by the mobile terminals and publish real-time road network information using web interface.The major contributions of this thesis are as follows:1) Building the cluster for dealing with mass transportation data based on the Spark framework, thus overcome the shortage of traditional cluster that can only deal with batch tasks. By building the distributed system in memory, the machine learning task that needs to iterate mass data can be finished in several seconds.2) Implementing the load balance strategy in the mobile cloud platform based on improved genetic algorithm. Using dynamic load balance to avoid hot spot and inefficient resource utilization.It overcomes the shortage of heterogeneity and dynamic computing resources in the mobile cloud environment and improves the resource utilization efficiency.3) Building high concurrency streaming cluster based on the reactor model. Overcoming the shortage of the traditional multi-thread model about synchronization, dead lock and thread switch cost. It increases the throughput of the cluster and improves the instantaneity of stream transport.