Exploring factors affecting learner satisfaction toward MOOCs : a hybrid approach of topic modeling, deep learning, and structural equation modeling /

Open in iSearch

Student Theses

2022

Chen, Xieling

Hong Kong : The Education University of Hong Kong

Thesis

The proliferation of Massive Open Online Courses (MOOCs) has created enormous research opportunities for understanding what drives a MOOC to succeed. However, few studies have exploited the ever-increasing amount of learner-generated course review data, especially by using deep learning, topic modeling, and Partial Least Squares-Structural Equation Modeling (PLS-SEM). Accordingly, based on 102,184 reviews from 401 MOOCs collected from Class Central and survey data collected from 168 university students in Mainland China, this study proposes a hybrid approach that combines these methodologies to understand learner satisfaction comprehensively. First, given that review's helpfulness is seldom considered in online education, this study explores machine learning's potential to classify MOOC review helpfulness automatically. Results revealed that the performance of machine learning was comparable to that of human coders. By emphasizing a machine's ability in the helpfulness of MOOC-review classification, this study contributes to a more accurate and efficient data analysis of reviews because a lack of review-helpfulness prediction may result in confusing insights. Second, given a lack of learner (dis)satisfaction research based on review topic modeling, this study examines factors leading to learners' dissatisfaction by considering review extremity, discipline differences, course grade levels, and topic correlations. Such analyses enable tailored insights into learner (dis)satisfaction in individual disciplines and MOOCs of different levels. The topic modeling result interpretation was guided by MOOC design theories and practices, which in turn, contributed to constructing a framework for MOOC design and evaluation. Unlike prior research based on generic learning design approaches or group-centered workshops, the framework was developed from learners' perceptions about learning. Third, given the lack of automatic MOOC review topic classification, this study applies deep learning to automatically classify MOOC reviews into ten topic categories to align with the constructed framework. Results showed recurrent convolutional neural network's superiority. This study contributes to MOOC-review analysis by significantly reducing coding workload and improving classification efficiency, thus enabling timely feedback about learners' concerns so providers can understand their MOOCs' performance. Fourth, to explore how the review topics affect learner satisfaction, this study uses PLSSEM to consider learners' sentiments hidden within textual review content. Results suggested that learner satisfaction was positively influenced by course content and structure, people involved, technology used, and learning processes. This study sets a precedent by adopting advanced PLS-SEM methodologies based on a large and unstructured online course review dataset to understand learner satisfaction automatically. Finally, to understand the fine-grained mechanism underlying learner satisfaction, this study conducts a survey study in Mainland China based on a theoretical model which considers MOOC design and implementation effects. This sets a precedent by investigating MOOC learners' satisfaction and continuance intention toward MOOCs based on the technology acceptance model, self-determination theory, and expectation-confirmation model. Results of PLS-SEM analysis showed perceived usefulness's positive effects on satisfaction and continuance intention. Assessment and autonomous motivation positively affected continuance intention; expectation confirmation significantly influenced perceived usefulness and satisfaction. This study provides new perspectives on resource allocation and course management strategies to promote learner satisfaction and retention

LG51.H43 Dr 2022eb Chenxl

https://educoll.lib.eduhk.hk/records/BsoNCkZY