| It is helpful to improve the relevance of the search results of search engines for people to search for valuable information quickly and efficiently in massive Web resources.Early researches mainly rely on the analysis of the query logs to infer users’ true query intentions.With the deepening of research,it is found that the information recorded in the query logs is limited,which is not conducive to the detailed analysis of users’ search behaviors.Eye Track Information(ETI)can really reflect users’ information concerns in the searching processes.However,it requires professional equipment to obtain ETI,which does not meet the actual application scenarios.Recently,researches have shown that there is a close relationship between mouse movement information(MMI)and ETI,so the MMI can be used for detailed analysis of users’ search behaviors.The main works done as follows:(1)This thesis proposes a Shapelets mining acceleration algorithm based on"key point"---KP-Shapelets.As for the problems and shortcomings in the current researches,this study introduces the time series data mining algorithm based on Shapelets to solve the problems of cursor track patterns mining,and proposes a Shapelets mining acceleration algorithm based on "key point".In this thesis,the accelerated KP-Shapelets mining algorithm is applied to the cursor track pattern extraction,replacing the original motifs-based cursor track pattern mining algorithm,which solves the problems of computational efficiency and users’ individual characteristics of motifs mining algorithm.(2)After studying the characteristics of different cursor track patterns,this thesis presents an analysis method of "split cursor track" pattern.Through in-depth study,we find that dividing a cursor track sequence into two one-dimensional time series can improve the distinguishing degree of cursor track pattern.Combining these characteristics of higher distinguishing degree with the KP-Shapelets mining algorithm can elegantly solve the problems of cursor track pattern classification.(3)Combining with KP-Shapelets mining algorithm,this study designs a personalized clustering application based on users’ cursor track pattern.It also verifies the availability and effectiveness of related work at the application level. |
