Penerapan Metode Ensemble Learning untuk Analisis Sentimen pada Data Twitter

Nasty, Khairuddin and Abatony, Nuzul (2024) Penerapan Metode Ensemble Learning untuk Analisis Sentimen pada Data Twitter. Project Report. [s.n], [s.l.]. (Unpublished)

Text 5025201041_5025201107-Project_Report.pdf - Accepted Version Restricted to Repository staff only Download (4MB) | Request a copy

Abstract

Penelitian ini bertujuan untuk menganalisis sentimen masyarakat terkait kenaikan harga Bahan Bakar Minyak (BBM) dan pemilu menggunakan metode ensemble learning yang menggabungkan model Machine Learning dan Deep Learning. Dataset diperoleh dari platform media sosial twitter yang berisi opini masyarakat mengenai kedua topik tersebut. Berbagai teknik vektorisasi teks (TF-IDF, Bag of Words, dan Word2Vec) digunakan untuk mengubah data teks menjadi format yang dapat dipahami komputer. Metode ensemble learning (stacking dan voting) diterapkan dengan menggabungkan model-model Decision Tree, SVM, Logistic Regression, KNN, serta model Deep Learning seperti LSTM, RNN, CNN, dan BiLSTM. Hasil penelitian menunjukkan bahwa pada dataset BBM, model SVM mencapai akurasi tertinggi 0,88836 (TF-IDF 1) dan LSTM mencapai 0,88235 (Word2Vec SkipGram). Pada dataset pemilu, LSTM unggul dengan akurasi 0,76044 (Word2Vec SkipGram), melampaui SVM yang mencapai 0,7191 (TF-IDF 2). Metode ensemble learning dengan Stacking LR optimal pada dataset BBM (akurasi 0,88636 dengan BOW), sedangkan Stacking SVM unggul pada dataset pemilu (akurasi 0,740770 dengan TF-IDF 2). Teknik vektorisasi TF-IDF meningkatkan akurasi model Machine Learning sebesar 2-5%, sementara Word2Vec meningkatkan akurasi model Deep Learning hingga 3-7%. Secara keseluruhan, kombinasi model melalui ensemble learning mengoptimalkan akurasi hingga 88,64% (BBM) dan 74,08% (pemilu), mengungguli performa model individu yang rata-rata mencapai 84-86% (BBM) dan 69-71% (pemilu).
============================================================================================================================
This study aims to analyze public sentiment regarding the increase in fuel prices (BBM) and elections using ensemble learning methods that combine Machine Learning and Deep Learning models. The dataset was collected from the social media platform Twitter, containing public opinions on these two topics. Various text vectorization techniques (TF-IDF, Bag of Words, and Word2Vec) were employed to transform textual data into a machine-readable format. Ensemble learning methods (stacking and voting) were applied by combining models such as Decision Tree, SVM, Logistic Regression, KNN, and Deep Learning models like LSTM, RNN, CNN, and BiLSTM. The results indicate that for the BBM dataset, the SVM model achieved the highest accuracy of 0.88836 (TF-IDF 1), and the LSTM model achieved 0.88235 (Word2Vec SkipGram). For the election dataset, LSTM outperformed others with an accuracy of 0.76044 (Word2Vec SkipGram), surpassing SVM, which achieved 0.7191 (TF-IDF 2). The ensemble learning method with Stacking LR was optimal for the BBM dataset (accuracy of 0.88636 with BOW), while Stacking SVM excelled for the election dataset (accuracy of 0.740770 with TF-IDF 2). TF-IDF vectorization techniques improved the accuracy of Machine Learning models by 2-5%, while Word2Vec enhanced the accuracy of Deep Learning models by 3-7%. Overall, the combination of models through ensemble learning optimized accuracy to 88.64% (BBM) and 74.08% (elections), outperforming individual models, which averaged 84-86% (BBM) and 69-71% (elections).

Item Type:	Monograph (Project Report)
Uncontrolled Keywords:	Analisis Sentimen, Deep Learning, Ensemble Learning, Kenaikan Harga BBM, Machine Learning, Model Klasifikasi, Pemilu, Stacking, TF-IDF, Voting, Word2Vec.
Subjects:	T Technology > T Technology (General) > T57.5 Data Processing
Divisions:	Faculty of Information and Communication Technology > Informatics > 55201-(S1) Undergraduate Thesis
Depositing User:	Khairuddin Nasty
Date Deposited:	09 Jan 2025 05:58
Last Modified:	09 Jan 2025 05:58
URI:	http://repository.its.ac.id/id/eprint/116228

Actions (login required)

View Item