Research And Implementation Of Intelligent Travel Platform Based On Big Data

With the development of cities and the increase of the number of trips per capita,the problem of traffic congestion is becoming more and more serious.Although at this stage have a large number of mature transportation system to alleviate traffic problems,but along with the arrival of the traffic data,the traditional transportation system for handling large amounts of heterogeneous data,in both storage and query and processing have existed bottleneck,based on this,this thesis designed and developed based on the wisdom of the big data travel platform,make better service in the field of intelligence traffic.In view of the shortcomings of the traditional transportation system,the main work of this thesis is mainly carried out from four aspects of system performance,data sharing,query optimization and function research and development,as follows:(1)In response to the large increase in public transport data,design its big data platform with the help of Hadoop,Spark,Hbase,Flume and other technologies to complete the collection and processing of heterogeneous data;according to the characteristics and usage of various types of data differently,the design uses Redis to store real-time data,the non-relational database Hbase and the distributed file system HDFS to store offline data;the Kafka message queue is used to eliminate the imbalance of upper and lower message processing,ensuring the stability and robustness of the system.(2)Using the front-end extraction scheme,the data of each business system is extracted into their respective front-end machines in a compressed format,and the data sharing and exchange between the various business systems are completed through the method of publishing and subscribing.(3)On the platform in the data warehouse construction scheduling tasks to execute slowest problems,this thesis proposes using the Spark SQL instead of the previous Hive SQL and partition optimization solution,and their performance comparison experiment,determined the final technology selection,and the experiment results show that the proposed optimization scheme can shorten the execution time of task scheduling,effectively improve the query efficiency.(4)Based on the study of factors affecting bus passenger flow,this thesis designed a passenger flow prediction module based on neural network under Hadoopplatform,and made use of experiments to show that the algorithm made up for the low efficiency under the parallel processing of Map Reduce,improved the timeliness of passenger flow prediction,and met the demand for public traffic big data processing.It has been proved by practice that this platform meets the needs of the public for green mobility services and the scientific management,monitoring,and dispatching of public transportation vehicles by enterprises and governments.The research and development of public travel platforms have a good guiding role.