Analysing machine learning models to detect disaster events using social media
Disasters are instabilities that occur on the interface between society and the environment. During disasters, people communicate to inform and request for support for themselves or their community. Social media is used as a medium for communication due to its wide reach and global audience. Duri...
Saved in:
Main Author: | |
---|---|
Format: | text::Thesis |
Language: | English |
Published: |
2023
|
Subjects: | |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
id |
my.uniten.dspace-19506 |
---|---|
record_format |
dspace |
spelling |
my.uniten.dspace-195062023-12-08T10:27:50Z Analysing machine learning models to detect disaster events using social media Faris Azni Azlan, Mr. Detect Disaster Events Disasters are instabilities that occur on the interface between society and the environment. During disasters, people communicate to inform and request for support for themselves or their community. Social media is used as a medium for communication due to its wide reach and global audience. During disasters, people communicate via messages regarding similar or different types of emergencies in the same general location. Interpreting and validating these messages during the occurrence of a disaster costs a significant time and loss. Therefore, this study presents a comparison between three algorithms, K-Nearest Neighbour (KNN), Naive Bayes (NB), and Support Vector Machine (SVM), to classify and sort messages so that the process of examining them can be simplified and accelerated. To simulate the examining process further, a fuzzy algorithm is developed to automatically rate the severity of a disaster as described in each message in disaster environment. The results are gauged using four statistics-based metrics and a time constraint. The statistics-based metrics are accuracy, precision, recall and f1-score. For accuracy, KNN and SVM tied with a score 0.79 or 79%. The same trend extends for precision, recall and f1-score, where KNN and SVM are equal in performance. In time constraint results, KNN is faster than SVM at producing output but is slower than NB. Despite being fastest among the three at producing output, NB has the lowest scores in the statistics portion of the evaluation results. The study has found KNN to be the most suitable algorithm to sort messages in a disaster situation for having the highest ratio of accuracy and speed out of the three models. 2023-05-03T13:35:26Z 2023-05-03T13:35:26Z 2020-07 Resource Types::text::Thesis https://irepository.uniten.edu.my/handle/123456789/19506 en application/pdf |
institution |
Universiti Tenaga Nasional |
building |
UNITEN Library |
collection |
Institutional Repository |
continent |
Asia |
country |
Malaysia |
content_provider |
Universiti Tenaga Nasional |
content_source |
UNITEN Institutional Repository |
url_provider |
http://dspace.uniten.edu.my/ |
language |
English |
topic |
Detect Disaster Events |
spellingShingle |
Detect Disaster Events Faris Azni Azlan, Mr. Analysing machine learning models to detect disaster events using social media |
description |
Disasters are instabilities that occur on the interface between society and the
environment. During disasters, people communicate to inform and request for support
for themselves or their community. Social media is used as a medium for
communication due to its wide reach and global audience. During disasters, people
communicate via messages regarding similar or different types of emergencies in the
same general location. Interpreting and validating these messages during the occurrence
of a disaster costs a significant time and loss. Therefore, this study presents a
comparison between three algorithms, K-Nearest Neighbour (KNN), Naive Bayes
(NB), and Support Vector Machine (SVM), to classify and sort messages so that the
process of examining them can be simplified and accelerated. To simulate the
examining process further, a fuzzy algorithm is developed to automatically rate the
severity of a disaster as described in each message in disaster environment. The results
are gauged using four statistics-based metrics and a time constraint. The statistics-based
metrics are accuracy, precision, recall and f1-score. For accuracy, KNN and SVM tied
with a score 0.79 or 79%. The same trend extends for precision, recall and f1-score,
where KNN and SVM are equal in performance. In time constraint results, KNN is
faster than SVM at producing output but is slower than NB. Despite being fastest among
the three at producing output, NB has the lowest scores in the statistics portion of the
evaluation results. The study has found KNN to be the most suitable algorithm to sort
messages in a disaster situation for having the highest ratio of accuracy and speed out
of the three models. |
format |
Resource Types::text::Thesis |
author |
Faris Azni Azlan, Mr. |
author_facet |
Faris Azni Azlan, Mr. |
author_sort |
Faris Azni Azlan, Mr. |
title |
Analysing machine learning models to detect disaster events using social media |
title_short |
Analysing machine learning models to detect disaster events using social media |
title_full |
Analysing machine learning models to detect disaster events using social media |
title_fullStr |
Analysing machine learning models to detect disaster events using social media |
title_full_unstemmed |
Analysing machine learning models to detect disaster events using social media |
title_sort |
analysing machine learning models to detect disaster events using social media |
publishDate |
2023 |
_version_ |
1806427982180384768 |
score |
13.18916 |