Implementasi Sistem Prediksi Saham IDX80 Berbasis Berita Ekonomi Berbahasa Indonesia Menggunakan LSTM dan Framework Streamlit

Abstract

The development of the Indonesian capital market, particularly for stocks included in the IDX80 index, generates a large volume of economic news information, making manual sentiment analysis challenging for investors. This study aims to build a sentiment classification system for Indonesian-language economic news articles related to IDX80 stocks using the Long Short-Term Memory (LSTM) method. The dataset consists of 3,000 articles obtained from trusted news portals. The research follows stages from data understanding to deployment. Text preprocessing includes case folding, cleansing, stopword removal, Sastrawi-based stemming, label encoding, tokenizing, and padding. The model architecture utilizes a 128-dimensional embedding layer, 128 LSTM units, 0.5 dropout, and Softmax activation. Experimental results show that the model achieved an accuracy of 92.67% in classifying positive, neutral, and negative sentiments. The model was evaluated using precision, recall, F1-score, and a confusion matrix. Furthermore, it was successfully implemented in a Streamlit-based application featuring single-article and batch CSV predictions. This system serves as a digital tool to assist investors in making objective investment decisions amidst market uncertainty.