The Research And Application Of Sequential Pattern Mining In RFID Data

It is more and more important to discover the path frequent pattern hidden in the path data generated in the RFID(Radio Frequency Identification)system operation process.The traditional methods based on the candidate generate and test need scan the sequence data set many times and can't fulfill the need of actual situation.Although the algorithm based on the pattern growth can reduce the scanning times,it will produce a large number of the projecting data set when facing the long path sequence pattern and dense data set,and in addition,there are many repeated patterns in the projecting data set.Moreover,there are some redundancies in the mining result of sequence pattern.As a result,the above defects bring the performance degradation of the traditional algorithms.Therefore,it is necessary to adapt the traditional sequence pattern mining technique to efficiently find the sequence patterns which hided in the RFID path data and to support the improvement of the management efficiency.This thesis focuses on the effective mining method to find the RFID path data's frequent sequence and the main achieving work as follows:Firstly,we give the bird view of the basic concept,the feature of the RFID path data and the sequence pattern mining techniques.Secondly,based on the concept of the closure sequence pattern,we propose the compact closed frequent sequence prefix tree structure and the efficient dynamic pruning technique which used in the PrefixSpan algorithm and present the technique which can extract the frequent sequence pattern from the path data set.The experimental results show the effectiveness of the technique.Thirdly,combining with the data flow's mining technique and adopting the sliding time window mechanism,we introduce the feedback technique to the dynamic time window mechanism and the time decay model for the history data to meet the requirements of the real RFID system.Furthermore,based on the incremental sequence pattern mining technique and the compact closed frequent sequence prefix tree structure,we give the sequence pattern mining algorithm which can be used in the data stream environment.Fourthly,the experimental results based on the synthetic dataset prove the effectiveness of the techniques proposed.