【作者】 刘杰；

【导师】 杨娟；

【作者基本信息】 四川师范大学 ， 教育技术学， 2019， 硕士

【摘要】 虽然在线学习系统可以有效地帮助学习者摆脱时间、地点等条件的限制进行学习,但是由于学习者在学习风格、学习偏向、学习优势等方面具有个体差异,因此会导致学习者产生不同的学习行为模式,而这些不同的学习行为模式可能会影响最终学习效果。目前的在线学习平台的学习过程数据采集因为缺乏足够的特征点,使得学习者的在线学习行为数据同质化严重,不仅无法有效的与最终学习效果关联起来,而且还会导致学习效果的不可预测。为了更好的发现学习者在学习过程中可能存在的问题以及这些学习行为对学习效果的影响,本文通过研究SPOC平台上的在线学习数据,得出了在线学习平台会因为提供的“无差别”服务会导致学习者产生知识建构的差异,而且单一的用户学习行为数据标签是无法准确预测学习者的学习效果的结论。该结论意味着采集在线学习行为差异化数据是非常有必要的,因此,本文利用Felder-Silverman学习风格模型作为采集差异化行为数据的依据,采集了4维的学习行为数据,采用粒子群算法优化后的贝叶斯网络对学习者阶段性学习效果进行预测。实验结果显示在获得部分学习行为数据的前提下,本文设计的预测评估模型能获得对学习者学习效果较好的预测效果,为智能学习系统可提前预判学习者学习风险并提供学习干预措施成为了可能。实验的结果同时也解释了由学习风格偏向性带来的学习行为差异以及学习行为与不同知识点之间的关联,为进一步设计更智能的在线学习平台提供了依据。更多还原

【Abstract】 Although online learning systems can effectively help learners to learn anytime and anywhere,individual differences such as learning preference,learning style and cognitive ability would lead to various online learning behaviors,as well as different learning effects.Many existing researches about how different learning behavours would influence learning effects still focused on the impacts that controlled factors may produce on the learning effects,few of them have noticed that those controlled factors may be a reflection of the behaviour features occurred during their learning processes,and can be represented by the key features that are recognized and extracted from the basic feature space which is used to describe the learning behaviours when learners are using the online platforms.In order to build a feasible feature space for the learning behaviours to describe individual differences,an analysis about how online learning behaviour may generate various online learning patterns that may influence learning effects is conducted based on a SPOC platform.The research reveals that by using learning style as the classification index,different learning patterns regarding different kinds of learning resources were adopted by different LS biased learners,and the difference between two kinds of learners was also appeared in their final learning effects and period learning effects.The research suggested that,the feature-space construction should be tightly connected with the users’ choices of the media kinds of the learning resources;besides that,the general intelligence theory of the knowledge integration suggested that the applying of the self-reflected information also should be considered in the feature space as one of the features that can be used to label a learning behaviour.Based on the multi-faceted learning behaviours that were collected from a prototype learning system in which a ‘C programming’ course is embedded,this paper proposed a learning effect regression/prediction model named FLR based on naive Bayesian model,and applied particle swarm algorithm to optimize the goal function;FLR not only significantly improved the prediction accuracy of the final learning effects compared with the linear regression with multi-variable,but also capable of explain the cause-result relations among learning behaviours and their learning effects(including learning effects in different learning phase).更多还原