How Predictable Are the Learning Data? Making Sense of Student Learning Behavior

• In Event: Innovative Learning Design in an Information and Communications Technology-Rich Environment: The Ongoing Story of Shanghai

Abstract

Objectives: This study investigates the extent to which student learning activities may accurately predict their academic achievements. A complex algorithm is proposed and a visualization tool is provided to trace the learning progress of students and identify the key indicators that need intervention.
Theoretical framework: Monitoring student learning activity is an essential component of quality education, and is a major predictor of effective teaching (Dawson et al., 2010; Siemens et al., 2011). Previous studies suggest that online learning data could offer a more accurate representation of student engagement in a course (Campbell et al., 2006). The findings can also be used to discern the pattern of the data, predict what should come next, and adopt the appropriate action to help students to learn more effectively (Chellatamilan et al., 2011).
Methods: The quantitative methodology comprises the primary source of data collection for this study. An experiment is to be conducted in the math course of the third grade at a primary school in Shanghai. Data will be tracked for the next semester. Four classes of 160 students and two teachers have been recruited to participate in the study. The context is a typical ICT-rich learning environment, with every student using a notebook as their “e-schoolbag,” and an e-Textbook is going to be used in the math class. Variables that demonstrate a significant correlation with student final grade will be identified. A best-fit predictive model for this course with regression modeling will be generated, and a visualization tool will be used to identify key indicators that need timely pedagogical intervention.
Data sources: Quantitative data will be derived from both learner off-put and intelligent data with the e-Textbook during the semester. The system can capture their inputs and collect evidence of their problem-solving sequences, knowledge, and strategy use, as reflected by the information each student selects or inputs. All collected data are categorized into four categories, namely, data about the course, the students, the resources, and access times. Qualitative data were derived from the content of interviews with students and teachers.
Preliminary findings: The established predictive model comprises key variables correlating significantly with student achievement, such as the number of attempts the student makes, the number of hints and feedback given, and the time allocation across parts of the problem. The developed visualization tool enables students and the instructor to easily understand the interpretation of the analysis results. The study is on-going, and a round of experiments will be conducted in September 2013. We would collect more samples and data to validate our model and hypotheses.
Significance of the study: The study will build knowledge about learning analytics in an ICT-rich environment in which the explicit and implicit learning behaviors of students can both be traced and analyzed. The results of the analysis can help students to obtain a better understanding of how their engagement relates to their success, and can assist teachers to better adapt their teaching and intervention based on student performance.

Authors

• Yiling Hu, East China Normal University

• Xiaoqing Gu, East China Normal University