Study Of Applying Sequential Pattern Mining To Highway Tunnel Traffic

With highway sharply developed, more and more tunnels of highway have been built. In order to discover accident potential that occurs in highway tunnel in time and try the best to decrease the negative influence of accident, Tunnel Monitoring System has been established. However, it uses real-time monitoring mechanism which lacks of testing data collation and analysis, resulting in a large amount of data being lost or useless. At the same time, lots of data have the relevance of time or sequence. Sequential pattern mining is to mining patterns that are frequent relative to time or other patterns in sequence database. As one of the most important tasks of data mining, Sequential pattern mining will have broad application in future such as the analysis of customer behavior, Web usage log analysis, DNA analysis and nature disaster prediction, etc.In this paper, based on the problems of tunnel monitoring system, the inadequacy in existing traffic detection technology is analyzed, tunnel traffic incident definition and grade has been discussed, a forecasting system of highway tunnel incidents based on sequential pattern mining is presented. Then considering the features of tunnel traffic data, a series of data pretreatment approaches is introduced. To provide high quality data for data mining, the technology of discretization is applied in original data. The main algorithms of sequential pattern mining are described, the application of PrefixSpan algorithm in tunnel traffic is researched, the pattern generated reflects the sequence features of tunnel traffic accidents. At last, the application of sequential pattern in tunnel traffic is researched, the sequence rules of traffic incident grade are established by the pruning strategy. Experiment shows that the rules can be used in forecasting the grade of traffic incident and provide intellectual support.