Application And Research Of MOOCs User Reading Behavior-oriented Frequent Pattern Mining

The ELearning Course is a new network teaching mode which is different from the traditional education.It is the network implementation about the evaluation contents and activities of courses.In this paper,the MOOCs(Massive Open Online Courses)is the set of several social networks,digital resources and Elearning Course.MOOCs relatives to the traditional course,it's without the interaction between teacher and student,among classmates.Then how to improve the quality of Elearning Course and the framework of course becomes the big problem of this new education model.Therefore,uses the frequent pattern mining technique to obtain the most frequent reading behavior sequence on MOOCs,and these sequences could be expressed as the hidden features about the user reading behavior.Moreover,uses these maximal frequent itemsets to solve the problem about optimizing the framework of course and planing the best reading path.In this paper,we use data set from OpenClassrooms(https://openclassrooms.com)which is the leading platform in European to research user reading behavior in MOOCs.Based on the concept of Reading Session,defined and analyzed the potential user reading behavior.Firstly,selected the most important behavior features from a set of user reading behavior,and built the transaction database about reading behavior in SQLite database management system.Secondly,data pre-processing data which is related to reading session and collected from transaction database in eclipse platform,the number of them is 489385 in 6 months.Then,does data extraction,data cleaning,data integration and data reduce and so on.After those previous operations,built the sequence based on Partid,Sessionid and duration.Moreover,uses the FPMax algorithm which can obtain the maximal frequent itemsets contained the longer result set and more information,in order to mine the protential user reading behavior on MOOCs model programming by Java.Put it in detail,improved the structure of MFI-tree in FPMax.Presenting the new views about replacing the single value with tuple as the input for inserting the MFI-tree,and output sequences according to the order of Sessionid rather than support,it's innovative.Then,obtains the reading behavior sequences order by Sessionid and the tuple in each sequence could be repeatable.Make up the lack of frequent pattern mining about result sequence only could contain the single data item,any duplicate value,and output the itemsets ordering of support and so on.Therefore,frequent pattern mining would be more suitable for the user reading behavior research.Proposed the RS-FPMax algorithm innovative.Finally,according to these two kinds of maximal frequent itemsets,uses an interdisciplinary technique which is the sequence alignment applied for DNA detection in biological information.Uses Needleman-Wunsch algorithm to obtain the sequence alignment between success' s maximal frequent itemsets and the test user reading behavior's sequence,at the same time uses Smith-Waterman algorithm to obtain the sequence alignment between failure's maximal frequent itemsets and the test user reading behavior's sequence.Finally,according to the score for sequence alignment,classifies the user who hasn't finished a course into the final user type(success or failed),then guide them to the more correct reading path and adjust the structure of a course.