Deep learning approach for student performance prediction in e-learning
The data in E-learning is generated as a result of the students' interactions during the learning sessions while accessing files and learning resources in E-learning. The learning system collects a huge amount of student’s data, such as the registration records, assessment results, and interact...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2019
|
| Subjects: | |
| Online Access: | http://eprints.utm.my/id/eprint/96299/1/SalmaHusseinOmarMSC2019.pdf.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The data in E-learning is generated as a result of the students' interactions during the learning sessions while accessing files and learning resources in E-learning. The learning system collects a huge amount of student’s data, such as the registration records, assessment results, and interaction log activities. The enormous amount of data stored in educational databases can provide useful information if analyzed and processed. Predicting the performance of students engaging with the E-learning platform is crucial and has a great impact on both the educational institute and students. This research aims at predicting the performance of students based on Deep Learning (DL) approach using Convolutional Neural Network (CNN). CNN trains a classifier on data by passing learned features through different layers of hidden features. CNN comprises of various layers of which are convolutional layer, pooling layer, activation layer, and fully connected layer. The dataset used in the study is obtained from Open University in the United Kingdom and it has been published in UCI repository. After performing feature selection, 25 attributes have been identified as very significant for student performance prediction using CNN. The training parameters such as the learning rate, weight decay, and optimizer were used to improve the performance of the CNN classifier. In addition to that, the effect that the number of convolutional layers, number of nodes and number of epochs were investigated and compared to evaluate whether it affects the accuracy of the prediction of the classifier. A comparative study being done showed that CNN out-performed Decision Tree and Artificial Neural Network algorithms by giving 72% accuracy, while the other two algorithms obtained 60.46% accuracy and 63.13 %, respectively. In conclusion, the CNN technique proves to be able to predict student academic performance and achieve high accuracy than the other two techniques. For future works, it is suggested to increase the data size, to change the environmental set up from CPU to GPU in order to investigate the performance of deep learning using CNN in large educational data. |
|---|