Research And Application Based On Similarity Of Event Sequences

The development of the Internet,Artificial Intelligence and other scientific technologies is unprecedentedly active,and the amount of data generated every day increases exponentially.The complex massive data also brings some new problems,that is,how to efficiently and effectively obtain valuable information from these data.These requirements have led to the birth and development of data mining.Data mining has a wide range of research and application fields,including event sequences data mining.Event sequences exist widely in daily life and grow rapidly.Traditional data mining techniques cannot be used directly to process this type of data so that event sequences data mining and analysis have received widespread attention.Research on the similarity measurement of event sequences is an important aspect in the field of event sequences data mining.At present,there are considerable research results.However,because of the complex characteristics of event sequences such as multiple attributes,wide range of event types and different source fields,many typical sequence similarity measurement algorithms are difficult to play a role,so research on the similarity measurement of event sequences in different application fields has important theoretical and practical value.This thesis takes the event sequences as the research object,and discusses in detail the shortcomings of the event sequences similarity in the field of user behavior research in the intelligent environment and travel route recommendation.In the field of user behavior research in intelligent environment,there is a lack of temporal and spatial multi-granularity cognition.In the field of travel route recommendation,little consideration is given to the influence of the visit order and play time of different types of attractions in the travel sequence on user preferences.In view of the above shortcomings,the corresponding improvement methods are proposed respectively.Therefore,this thesis mainly includes the following work:1)In the intelligent environment,the existing similarity research on user behavior has rarely performed similarity analysis on behavioral events from different temporal granularity,spatial granularity,and activity type granularity.It lacks a comprehensive understanding of user behavior with multiple granularities and angles.The Multi-granular Spatiotemporal Sequences Alignment algorithm is designed,and the user behavior event sequences are compared from different granularities through granularity control.Finally,the experiments verify the effectiveness and efficiency of the algorithm.2)When modeling and solving travel route recommendation problems based on user preferences,existing research on travel route recommendations does not consider the order in which users visit attractions,the type of each attraction and the length of time spent at each attraction.This thesis comprehensively considers the impact of different types of attractions,play time and the order of visiting attractions on user preferences,and designs a travel route recommendation framework based on MGSSA algorithm.Finally,the effectiveness and robustness of the method are experimentally evaluated.Based on this framework,a personalized travel route recommendation system is implemented.