Prediksi Harga Saham Berdasarkan Sentimen Publik Atas Layanan Telekomunikasi Menggunakan Pendekatan Gated Recurrent Unit

Abstract

Stock market serves as a highly popular investment instrument in Indonesia, influenced by various factors, including public sentiment towards telecommunication services. This research aims to analyze and predict the movement of Telkom's stock prices based on public sentiment on the Twitter platform, employing a deep learning approach utilizing the Gated Recurrent Unit (GRU). The Twitter data used specifically includes tweets referring to Telkom's services ($TLKM.JK), while historical stock data from Yahoo Finance is utilized as a supporting dataset. Sentiment Analysis is conducted using VADER to classify sentiments into positive, negative, or neutral categories. The sentiment data is split with 80% for the Training process and 20% for model testing. In contrast to previous studies using LSTM models and Reporting an RMSE of 1120.6517, the findings of this research indicate that the GRU model can predict Telkom's stock prices with an accuracy level reaching 90%. The evaluation results of this model show an MSE of 102.43 and an RMSE of 10.120770.