Abstract
Learners' satisfaction with Massive Open Online Courses (MOOCs) has been evaluated through quantitative approaches focusing on survey-based methods in several studies. User-Generated Content (UGC) has been an effective approach to assess users' interactions with e-learning systems. Other than survey-based methods, the UGC generated from MOOCs learners can provide wider perspectives of learners' experiences using text mining approaches. Educational Data Mining (EDM) uses data mining, machine learning, and statistics to explore the information generated from educational portals. This study aims to explore learners' levels of satisfaction with MOOCs by presenting a new hybrid approach for EDM that combines both machine learning and survey-based methodologies to investigate the factors that can enhance learners' satisfaction with MOOCs. To address the goal of this study, the Latent Dirichlet Allocation (LDA) is used for analyzing learners' reviews, Self-Organizing Maps (SOM) is used for data segmentation, and Adaptive Neuro-Fuzzy Inference Systems (ANFIS) is utilized for predicting the learners' satisfaction from the identified factors. Based on the analysis of the first stage using the EDM approach, a research model is designed, and a questionnaire is distributed among MOOCs learners. The data is analyzed using PLS-SEM to provide proof of the reliability and validity of the research model and to confirm the significance of the research paths. Several methodological and practical contributions are presented based on the analysis of the proposed methodology.