Course-tailored Prediction Model: Enhancing Learner Performance Prediction by Adaptive Use of Proxy Variables

• In Event: KSET - Learning analytics

Short Description

We (a) presented a data mining process to construct proxy variables indicative of learners' high performance in asynchronous online discussion (AOD) contexts, (b) compared the accuracy of local prediction models to that of generic prediction models, and (c) proposed an adaptive prediction system (APS) that generates local prediction models. The result indicates: (1) a local prediction model outperforms a generic model in terms of accuracy and stability, and (2) the proxy variables are valid predictors that represent indicators of successful learning in AOD.

Authors

• Meehyun Yoon, The University of Georgia

• Dongho Kim, The University of Georgia

• Yeonjeong Park, Ewha Womans University

• Il-Hyun Jo, Ewha Womans University