Designing an E-Repository of Sentiment Data and Cyberbullying Detection in Indonesian using a Parameter Optimization Algorithm for LSTM

Bagioyuwono, Michael Abhinaya and Kristiyanti, Dinar Ajeng and Nugroho, Antonius Sony Eko (2024) Designing an E-Repository of Sentiment Data and Cyberbullying Detection in Indonesian using a Parameter Optimization Algorithm for LSTM. International Journal of Intelligent Systems and Applications in Engineering (IJISAE), 12 (3). pp. 4093-4106. ISSN 21476799

Text Designing_an_E-Repository_of_Sentiment_Data_and_Cyberbullying.pdf Download (2MB)

Abstract

The rise in the number of social media users, particularly in Indonesia, has led to an increase in the prevalence of cyberbullying cases in Indonesia. One of the rising social media platform twitter is nominated as the moxt toxic platform. One approach to preventing cyberbullying on social media is to analyse a person's opinion or assessment of the sentiment or emotion expressed on the platform. Sentiment Analysis, has been shown to be an effective strategy for identifying and addressing cyberbullying. Hyperparameter tuning is crucial for enhancing the performance of deep learning models like Long Short-Term Memory (LSTM), which often struggle with optimizing parameters due to local minima. This challenge is tackled using Particle Swarm Optimization (PSO) and Salp Swarm Algorithm (SSA) for more effective tuning.. The research encompasses various stages: data scraping, preprocessing, translating, labeling, and modeling with an LSTM optimized by PSO and SSA to determine the optimal number of LSTM units. This is followed by statistical testing and evaluation. The optimal model will be utilized in the data repository website and cyberbullying classification based on user input and allows users to download and upload datasets with administrator permission. The finding shows that models LSTM that been optimised with PSO (PSO-LSTM), has the best performance between the conventional model and the SSA-LSTM model. The PSO-LSTM algorithm produces 87.43% accuracy, 41.29% loss, and 12.93 seconds execution time. Results of the website data repository design have been tested with the User Acceptance Test with the results running in accordance with the expected results.

Item Type:	Article
Keywords:	CyberbullyingDetection, Database Repository, Long Short-Term Memory (LSTM), Optimizing Parameters, Particle Swarm Optimization (PSO), Salp Swarm Algorithm (SSA)
Subjects:	000 Computer Science, Information and General Works > 000 Computer Science, Knowledge and Systems > 005 Computer Programming 000 Computer Science, Information and General Works > 000 Computer Science, Knowledge and Systems > 005 Computer Programming > 005.2 Programming for Specific Computers, Algorithm, HTML, PHP, java, C++
Divisions:	Faculty of Engineering & Informatics > Information System
Depositing User:	Administrator UMN Library
Date Deposited:	04 Sep 2024 09:03
Last Modified:	04 Sep 2024 09:03
URI:	https://kc.umn.ac.id/id/eprint/31211

Actions (login required)

View Item