A Hybrid Method Using Ensembles of Neural Network and Text Mining for Learner Satisfaction Analysis from Big Datasets in Online Learning Platform

Computational intelligence approaches have proven to be effective in enhancing online learning systems. Although many studies have been conducted to reveal the learners' satisfaction in online learning platforms, the use of machine learning in the analysis of big datasets for this aim has rarely been explored. In addition, although the analysis of online reviews on courses has been carried out in other fields, there are very few contributions in the area of online learning platforms. This study, therefore, aims to perform learner satisfaction analysis through the use of machine learning. We develop a new method using text mining and supervised learning techniques with the aid of the ensemble learning approach. A boosting approach, AdaBoost, is used in ANN for ensemble learning to improve its performance. We employ Artificial Neural Network (ANN) approach, dimensionality reduction and Latent Dirichlet Allocation (LDA) for textual data analysis. Principal Component Analysis (PCA) is used for data dimensionality reduction. We perform several experimental evaluations on the big datasets obtained from the online learning platforms. The accuracy and computation time of the proposed method are assessed on the obtained dataset. The method is compared with several machine learning approaches to show its effectiveness in big datasets analysis. The results showed that the method is effective in predicting learners' satisfaction from online reviews. In addition, the proposed method outperform other classifiers, K-Nearest Neighbor (K-NN), Decision Trees (DT), Support Vector Machines (SVM) and Naive Bayes (NB), in case of accuracy. The results are discussed and research implications from different perspectives are provided for future developments of educational decision support systems.