Research On Student Learning Behavior Analysis Based On Process Mining

With the popularity of various learning support environments and information systems,it is possible to accumulate massive online event log,such as continuous tapping data streams,chat records,text editing historical data,action tracking,learning resource usage records,etc.,process mining can monitor and prompt the learning process with these records.Process mining builds a bridge between data mining and business process modeling and analysis,which is a sub-discipline that adds a process-oriented perspective to data mining.Process mining is generally successful in discovering learning patterns in the learning environment,and the current relevant research has not fully explored the application potential of educational process mining,only to discover the learning process of students,and has not fully explored the hidden information contained in it,such as the hierarchical information of learning behavior,the duplicate of learning tasks,and lacks application examples of universal significance.This paper focuses on the above problems and expands the application potential of the educational process in real teaching scenarios,and the main research contents are as follows:(1)A method for identifying the educational process discovery of duplicate learning tasks is proposed,the specific idea is to assume that the context of repeated learning tasks is different,and the transition system is introduced to identify the execution of learning tasks,according to this characteristic,extract the transition system.In the same state,the sequential relationship between input transition and output is constructed,and a direct following that can describe the precursor and successor between duplicate learning tasks is constructed.Then the process tree is constructed with the help of an inductive discovery algorithm to mine more complex task relationships in the direct following relationship.(2)A discovery method of hierarchical learning behavior model is proposed,including the implementation method is to further discover the nested relationship between learning behaviors by analyzing the time information of online learning behaviors,and construct a behavior nesting relationship tree to describe the correspondence between the upper behavior and its lower behavior,so as to create a hierarchical learning behavior event log according to the behavior nesting relationship tree.Finally,find the upper learning behavior model and the lower learning behavior model called by each upper level discovered through the hierarchical event log.(3)The remaining time prediction method and application frame of the education process are proposed,the specific implementation steps are to improve the granularity of the student learning behavior event log to improve the prediction accuracy of the prediction model,and then intercept the prefix construction set of the learning trace as the input of the prediction model.Finally,call the recurrent neural network to predict the remaining learning time of the students.Then application architecture of the education process takes the real-time student trajectory as the input to recommends the most suitable learning trajectory in the set of excellent student trace according to the sequence of real-time trace.