(1119) DESIGN OF DIGITAL TEXTBOOK-BASED LEARNING ANALYSIS SYSTEM FOR PERSONALIZED EDUCATION

摘要

In South Korea, classes are currently using digital textbooks in pilot schools. Digital textbooks areopen media which not only provide learning content and materials in a flexible manner according toindividual learning levels and styles but also enable easy connection to external third-party tools, usingvarious devices (mobile, PC, etc.) and cloud computing environment.In the conventional educational field, most data related to learning activity disappear when educationis over. However, in digital textbook-based class, various forms of learning activity data can becollected, used, and analyzed thanks to the use of on-line lesson tools. In this background, wecommenced this research to improve the learning environment through analysis of data stored in thisway.IMS Global proposes the Caliper Framework[1] to measure and analyze student achievements. TheCaliper Framework has the structure of collecting data generated mostly from online learning system,which is central to e-Learning activity like LMS, through the standard Sensor API and assessing thedata through metrics that can measure the data. However, this framework is limited in its application toKorea's K12 digital textbook-based education environment: In this environment, we should collect andanalyze learning activity data from classes using online learning tools that are diversely dispersed inclassroom-centered, not online-centered, lessons.Therefore, we need to develop an analysis platform which will enable comprehensive assessment ofvarious forms of data generated from in-class educational process, including the use of digitaltextbooks, participation in learning communities, digital textbook content, and evaluation by teachers.Our research includes methods of collecting and storing data for analysis of teaching and learningachievements from these various and unprepared data for analysis. We design and implement asystem that collects, analyzes, and visualizes data through stages of data collector, pre-processor(three-level metric generator), analyzer, and visualizer.In this paper, we describe methods of extracting metric data that can be used for analysis from variousforms of data generated without considering learning analysis in the pre-processing phase. We dividedthis process into three levels in a systematic manner.The first level metric is data collectable through queries of data to be collected. We add a timestamp tonon-standard data collected from diverse systems if necessary and accumulate the data in mongoDBas log data. The first level metric refers to data set that can be generated through queries of theaccumulated log data.The second level metric refers to data that can be generated by applying simple forms of statisticalfunctions to data set gained as the result of the first level metric. For example, such statisticalfunctions as total sum, count, average, and min/max value can be used.The third level metric refers to data set developed by applying multi-dimensional data conversion anddata mining techniques from the data extracted not just from the first and second level metrics but alsofrom other third level metric. Algorithm that can be used in extracting the third level metric may includesimilarity analysis through duration-based pattern matching and data mining. These third level metricdata can be used to analyze whether students have planned patterns of learning activity based ontheir plans and self-directed learning skills.