Emerging Sequences Mining Based On Location Information

With the development of information technology,sequence information filled our life every moment,sequence information in each area also presented the explosive growth.Under the background of information age,how to efficiently find the useful information from the sequence information,become a great challenge for the sequential pattern mining.Emerging sequence,with purpose of finding the sub sequence pattern with strong ability of differentiation,and then to build an efficient classifier.In the past,mining of sequential pattern mining is based on the support degree or the number of feature selection framework,which only focused on the existence of the emerging sequence,and ignored the other information that the sequence itself carries.With the increase of the amount and dimension of information,the traditional algorithm is stretched in dealing with more and more demanding requirements.Therefore,many researchers began to try to find a more effective way to reveal sequence patterns mining.In this paper,a new emerging sequence has been proposed,named location emerging sequences,short for LESs.Based on the number of occurrences,a new sequence pattern mining framework with location has been constructed,and then design the related algorithm to mine location emerging sequence.It not only pays attention to the pattern of the number of occurrences,but also integrated the location information carried by the model itself.Take a combination with location and the number of pattern,it is important to study emerging sequence pattern mining.The main work and contributions are as follows:(1)For problem that traditional emerging sequential pattern mining algorithms ignore subsequence itself carries the information,we proposed location emerging sequence,the location information of subsequence has been introduced.Based on the number of occurrences of framework,a new sequence pattern mining framework with location has been constructs.The position information of the location sequence pattern is effectively combined.In order to effectively tap location emerging sequential patterns,we used the location information of the suffix tree to store the data,in a certain extent reduces the time complexity of the algorithm,and the algorithm can adapt to the higher dimensional data.Experiments show that the location information has an important influence on the classification performance,and the classification performance of the classifier constructed by the position exposure sequence is improved by about 2%compared with the previous algorithm.(2)How to reasonably introducing location information is the focus of this paper,according to the problem which the location of emerging sequential patterns mining can't adaptive adjust the position coefficient K and location emerging sequential pattern mining algorithm framework is not compact enough,we propose adaptive emerging sequential pattern mining algorithm,the location emerging sequential pattern mining more flexible,can adapt to the demand of data sets and the changes of user.In the new definition,the pattern of the location emerging sequence is redefined,and the number of times and the position information are combined closely,and the feature selection is carried out to reduce the complexity of the algorithm.The experimental results show that the pattern of the position exposure sequence has an important influence on the classification performance,and the number of the patterns of the position exposure sequence can't be ignored.