Understanding the effectiveness of NLP deep learning for covid-19 prediction in Indonesia | |
| Author | Sandri, Widya Eka |
| Call Number | AIT Thesis no.ICT-21-01 |
| Subject(s) | COVID-19 Pandemic, 2020---Indonesia Natural language processing (Computer science)--Indonesia Deep learning (Machine learning) |
| Note | A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Information and Communication Technologies |
| Publisher | Asian Institute of Technology |
| Abstract | Many COVID-19 prediction models have been implemented using various methods, rang ing from traditional epidemic models to advanced machine learning models. However, most predictive models proposed rely on COVID-19 case data but ignore textual factors that may affect the COVID-19 situation, such as government policy, public opinion, the current po litical situation, and general information. In this study, we propose prediction models that incorporate both COVID-19 cases and embedded textual information from Twitter using nat ural language processing (NLP) techniques (embeddings and topic modeling) and long short term memory (LSTM) to predict the number of COVID-19 new cases in the upcoming seven days. Generally, models that utilize both new cases and textual information show significant improvements in prediction performance compared to models without textual information. Best performing model uses textual features extracted using bidirectional encoder represen tations from transformers (BERT) with an average mean absolute error (MAE) of 573.82 and mean absolute percentage error (MAPE) of 9.66%. Other models, such as LSTM, linear regression (LR), and susceptible-exposed-infected-recovered (SEIR) had lower accuracy in prediction without textual information. Moreover, we find that information related to gov ernment policy substantially improves the COVID-19 prediction performance across all text embeddings. |
| Year | 2021 |
| Type | Thesis |
| School | School of Engineering and Technology (SET) |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Information and Communication Technology (ICT) |
| Chairperson(s) | Chaklam Silpasuwanchai |
| Examination Committee(s) | Dailey, Matthew N.;Attaphongse Taparugssanagorn |
| Scholarship Donor(s) | His Majesty the King’s Scholarships (Thailand) |
| Degree | Thesis (M. Sc.) - Asian Institute of Technology, 2021 |