SENTIMENT ANALYSIS OF CONVERSATION AND RADICAL VIEW NETWORKS ON TWITTER USING SUPERVISED MACHINE LEARNING
The background of this research is the phenomenon of self-radicalization which is internet mediated. in this way people can go through the entire radicalization process from candidates to carrying out terrorist operations. The narrative on radicalism becomes the end of the initial amplification...
Saved in:
| Main Author: | |
|---|---|
| Format: | Theses |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/54518 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | The background of this research is the phenomenon of self-radicalization
which is internet mediated. in this way people can go through the entire
radicalization process from candidates to carrying out terrorist operations. The
narrative on radicalism becomes the end of the initial amplification of the action.
This narrative can be threatening when acquired by people with identity crisis.
Aforesaid narrative, if it is poorly controlled, can be dangerous for sustainability
of nationality, even more dangerous than the terrorism itself.
This research attempts to analyze the sentiment of the radicalism discourse
phenomena as an organized, integrated, systematic, and sustainable process
carried out by the accounts that are related to the radicalism issue in the form of
Twitter post (tweet). This research also wants to describe the form of the Social
Network Analysis (SNA) network, the role and involvement of the Twitter accounts
related to radicalism to cut the communication lines and the amplification intensity
of their ideas.
We categorize radicalism discourse on Twitter in the form of sentiment clusters,
using the Supervised Machine Learning method with the naïve baiyes algorithm
and Nazief & Adriani stemming. We also analyzed the communication network
activity, intensity, role and involvement of the Twitter accounts in the radicalism
discourse, mapped using SNA by aggregating the Twitter accounts involved in the
conversation.
The scenario testing has been carried out where the training data and the
testing data taken from the 1,060 datasets with the percentage of distributions are
75% for the training data and 25% for the testing data. The testing with this
scenario produces the values of confusion matrix accuracy with a confidence level
of up to 92%. This is higher than previous studies with SVM, which is only 70% to
83%. The results of the analysis found that rejection of radicalism was represented
by negative sentiment (44.46%), a slight difference compared to the positive
sentiment (40.91%), representing a group that was skeptical of counter-narrative
of radicalism. Meanwhile, neutral sentiment (14.63%) represents the group who
are doubtful and apathetic towards the counter-narrative of radicalism. |
|---|