基于过程挖掘的分层学习行为模型发现方法

Discovery method of hierarchical learning behavior model based on process mining

教育过程挖掘将过程挖掘技术应用于教育数据分析,是教育数据挖掘的重要分支之一。当前教育数据挖掘主要是用经典的机器学习算法对在线学习数据进行建模分析,难以描述全局的学习过程。凭借解析事件日志发现控制流模型的过程挖掘技术可以解决这一难题,但由于真实数据受各种客观因素的影响,存在大量噪声和无关行为,已有的挖掘方法往往会生成“意大利面模型”,不利于分析理解。针对这一问题,本研究提出分层过程挖掘方法发现学生学习过程。具体方法是:通过解析带生命周期事件日志的时间属性,发现活动嵌套关系;然后构造分层事件日志,进而挖掘描述学习行为的分层过程模型;最后用契合度、精确度、F-值三个指标,系统地比较分层过程挖掘方法与已有过程挖掘方法所挖掘模型的区别。

Educational process mining applies the process mining technology to educational data analysis and it is an important branch of education data mining.Mainly using classic machine learning algorithms to establish models for the analysis of the student online learning data,the current education data mining is difficult to describe the global student learning process.This problem can be solved by the process mining technology used to discover the control flow model by analyzing event logs.But the existence of noise and irrelevant behavior sequences in real data affected by various factors tends to make the existing mining methods generate a“Spaghetti model”and thus are not conducive to analysis and understanding.In view of this problem,this paper proposed the Hierarchical Process Mining(HPM)algorithm to discover the student learning behavior process.Firstly,the time attributes of the life-cycle event logs were analyzed to discover the nesting relationship of activities.Then,the layered event logs were constructed and the hierarchical process model for describing the student learning behavior was mined.Finally,the indicators of fitness,precision,and F-scores were used to make a systematical comparison between the mining models established by the layered process mining method and the existing process mining methods.