Research Of Intelligent Transport Platform Based On Big Data

With the diversification of services and facilities to improve transportinfrastructure and the data acquisition system, the amount of data traffic system is in astate of rapid growth, the era of big data with intelligent transportation system has come.Transportation system is a real-time, changing, with the increase in the number oftraffic data，traditional ways of dealing data can not meet our needs, especiallyexcessive delay and inaccuracies in real-time traffic prediction and optimal route bootperformance is very poor areas, the use of cloud storage can become massive datastorage costs low while increasing data security, cloud computing can handle hugeamounts of data to achieve the most efficient use of available resources to quicklyprocess the data, so the use of cloud storage processing of massive data intelligenttransportation is inevitable.Among the lot of cloud processing platform, Hadoop platform with ITS entirety,the characteristics of the system into the mainstream open source cloud platform,Hadoop can take advantage of a lot of cheap server clusters to complete the fast parallelprocessing of large data. The core component element includes a Hadoop distributedfile system and MapReduce, Hadoop distributed file system to provide safe, reliableand efficient storage capabilities, MapReduce provides massive data parallelprocessing capabilities, which allows us to take advantage of Hadoop platform andefficient processing of intelligent transportation systems massive data miningproblems. While MapReduce framework has some drawbacks, such as the single pointof failure probability is high, low resource utilization, but it upgraded version YARNgood solution to this problem, so Hadoop platform for applications in intelligenttransportation system is very broad.Based on the analysis of the transport system, found that the more commonly usedfeatures real-time traffic and route guidance predicted correlation is relatively large, wecan use a method based on historical data to make predictions on the road and use theresults to predict the dynamic assignment of the road, and then the use of single-sourceshortest path algorithm to simulate the optimal path, although such a manner that thelarge increase in computing, such as dynamic assignment of all sections of the need forreal-time prediction method based on historical data, but we use Hadoop platform a good solution computational problems.Through analysis of the application of data mining in Hadoop ’s strengths, inreal-time traffic prediction system based on historical data, we use KNN algorithm fordata mining, real-time and historical traffic data model to predict the traffic matching,and then KNN algorithm use MapReduce computation model to achieve parallelism toreduce the data processing time, and most problematic path selection, we useMapReduce-based parallel single-source shortest path algorithm to achieve higherefficiency. Finally, the cloud platform application framework for intelligenttransportation systems analyzed design, the system is divided into four five parts, eachlayer independently to provide services to the top, in addition to more efficient storageof data processing, the use of cloud storage also using a conventional data storage datain order to achieve higher efficiency.Finally, KNN based MapReduce parallel algorithms to predict traffic andsingle-source shortest path algorithm using optimal paths in parallel simulation testscarried out by the comparison test to prove the use of intelligent transportation Hadoopclusters to enhance the performance of great help.