Analisis Sentimen Berita Pasar Saham Indonesia Menggunakan Metode Convolutional Neural Network.

Effendi, Yaniar Pradityas (2022) Analisis Sentimen Berita Pasar Saham Indonesia Menggunakan Metode Convolutional Neural Network. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 05111840000047-Undergraduate_Thesis.pdf Restricted to Repository staff only Download (2MB)

Abstract

Perubahan harga saham berhubungan dengan sentimen berita. Sebagai investor, penting untuk menganalisis informasi pasar termasuk berita pasar saham untuk menghindari risiko investasi dan memperoleh keuntungan yang lebih baik. Salah satu metode yang dapat digunakan untuk mengetahui sentimen dari berita pasar saham adalah analisis sentimen. Studi sebelumnya menunjukkan bahwa deep learning dianggap dapat menghasilkan akurasi model yang lebih tinggi daripada pendekatan machine learning biasa dan beberapa studi menunjukkan model Convolutional Neural Network (CNN) dianggap menghasilkan performa paling baik untuk klasifikasi teks. Untuk itu dibangun model analisis sentimen menggunakan metode CNN yang dikombinasikan dengan model word embedding menggunakan word2vec dan tanpa word2vec. Selain CNN, model dibandingkan pula dengan metode Long Short-Term Memory (LSTM) dan metode gabungan CNN-LSTM. Adapun data yang digunakan merupakan berita pasar saham hasil crawling dari portal berita Kontan, Bisnis Indonesia, dan CNBC Indonesia periode 15 November 2016 – 15 November 2021 sejumlah 3001 artikel yang dikelompokkan menjadi sentimen positif, negatif, atau netral. Tahapan dari tugas akhir ini adalah pengolahan data, pra-pemrosesan teks, ekstraksi fitur, training, dan testing model. Uji coba dilakukan dengan beberapa skenario metode klasifikasi dengan model word embedding berbeda serta uji coba kombinasi parameter tiap metode klasifikasi. Berdasarkan hasil uji coba, performa analisis sentimen terbaik adalah menggunakan metode CNN tanpa word2vec dengan nilai akurasi, rata-rata precision, rata-rata recall, dan rata-rata f-score sebesar 85,00%, 85,30%, 85,00% dan 85,10%.
=================================================================================================================================
Changes in stock prices are related to news sentiment. As an investor, it is important to analyze market information including stock market news to avoid investment risks and get better returns. One method that can be used to determine the sentiment of stock market news is sentiment analysis. Previous studies have shown that deep learning is considered to result in higher model accuracy than traditional machine learning approaches and several have shown that the Convolutional Neural Network (CNN) model produces deep learning models with the best performance for text classification. For this reason, a sentiment analysis model was built using the CNN method combined with a word embedding model using word2vec and without word2vec. In addition to CNN, the model is also compared with the Long Short-Term Memory (LSTM) method and the combined CNN-LSTM method. The data used is stock market news crawled from news portals Kontan, Bisnis Indonesia, and CNBC Indonesia for the period 15 November 2016 – 15 November 2021, a total of 3001 articles are grouped into positive, negative, or neutral sentiments. The stages of this final project are data processing, text pre-processing, feature extraction, training, and model testing. The trials were carried out with several scenarios of the classification method with different word embedding models as well as testing the combination of parameters for each classification method. Based on the test results, the best sentiment analysis performance is using the CNN method without word2vec with values of accuracy, average precision, average recall, and average f-score of 85.00%, 85.30%, 85.00 % and 85.10%.

Item Type:	Thesis (Other)
Additional Information:	RSIf 006.32 Eff a-1 2022
Uncontrolled Keywords:	Analisis sentimen, Convolutional Neural Network, Long Short-Term Memory, Word Embedding. Sentiment Analysis, Convolutional Neural Network, Long Short-Term Memory, Word Embedding.
Subjects:	Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Mr. Marsudiyana -
Date Deposited:	25 May 2026 06:44
Last Modified:	25 May 2026 06:44
URI:	http://repository.its.ac.id/id/eprint/133399

Actions (login required)

View Item