Research On Data Mining In MOOC Learning Behavior

With the continuous development of network and popularity of Internet applications, information technology are changing our traditional teaching mode of education. From 2012, Massive Open Online Course (MOOC) rose in the United States and has set off a worldwide educational technology revolution, which is welcomed by college teachers and students as well as the public generally.During the process of MOOC studying, learners has produced a great deal of data related to their learning behavior, and the big data can deeply reflect learners’ learning habits and learning style characteristics. Thus, digital learning, learning data mining and learning behavior analysis has attracted more and more researchers, which become an integrated research area combining education theory and computer application-the concepts of a computing education.This thesis focuses on the learning behavior data which learners produced in MOOC learning process, such as time of watching video. Different learners’ learning behaviors exist individual differences. Embarking from mining the learning behavior data of learners in MOOC, this thesis mainly work in the following several aspects:(1) Propose a Causal Association Analysis (CAA) Algorithm. From learners’ learning behavior in MOOC learning environment, this thesis gives the definition of learners’ learning behavior in MOOC learning environment, illustrates the components of learning behavior, and presents a formalized representation of learning behavior. Through improving Apriori algorithm, this thesis proposes causal correlation analysis (CAA) algorithm to analyze the association between learning behavior and effect, and also and points out an application direction of inspection system based on the learning behavior of daily learning on.(2) Propose Weighted K-means (WK-means) Algorithm based on CAA algorithm. After getting learning behavior that influence the learning effect, this thesis does a clustering analysis job on constituted eigenvector of several learning behavior and effect. To obtain better classification effect, this thesis uses weighted K-means algorithm, which gives weighted value, according to the result of CAA, to learning behavior. According to the result of clustering analysis, all learners are divided into four categories, and suggestions for learners, teachers as well as MOOC platform design are put forward.(3) Create Learning Style Model by Decision Tree. Different learning behaviors combination constitute the learners’ learning style, through the learners’ learning behavior combination, we can analyses the learner’s learning style, and recommend learning partners of appropriate style based on the classification of learning style, so as to better improve the learning efficiency.