PERBANDINGAN METODE LONG SHORT-TERM MEMORY (LSTM) DAN GATED RECURRENT UNIT (GRU) DENGAN PENDEKATAN UNIVARIAT DAN MULTIVARIAT UNTUK MEMPREDIKSI HARGA SAHAM (STUDI KASUS : HARGA SAHAM BMRI PERIODE 2015-2024)

Amanda Riyas Utami, NIM.: 22106010003 (2026) PERBANDINGAN METODE LONG SHORT-TERM MEMORY (LSTM) DAN GATED RECURRENT UNIT (GRU) DENGAN PENDEKATAN UNIVARIAT DAN MULTIVARIAT UNTUK MEMPREDIKSI HARGA SAHAM (STUDI KASUS : HARGA SAHAM BMRI PERIODE 2015-2024). Skripsi thesis, UIN SUNAN KALIJAGA YOGYAKARTA.

Preview	Text (PERBANDINGAN METODE LONG SHORT-TERM MEMORY (LSTM) DAN GATED RECURRENT UNIT (GRU) DENGAN PENDEKATAN UNIVARIAT DAN MULTIVARIAT UNTUK MEMPREDIKSI HARGA SAHAM (STUDI KASUS : HARGA SAHAM BMRI PERIODE 2015-2024)) 22106010003_BAB-I_IV-atau-V_DAFTAR-PUSTAKA.pdf - Published Version Download (8MB) | Preview
	Text (PERBANDINGAN METODE LONG SHORT-TERM MEMORY (LSTM) DAN GATED RECURRENT UNIT (GRU) DENGAN PENDEKATAN UNIVARIAT DAN MULTIVARIAT UNTUK MEMPREDIKSI HARGA SAHAM (STUDI KASUS : HARGA SAHAM BMRI PERIODE 2015-2024)) 22106010003_BAB-II_sampai_SEBELUM-BAB-TERAKHIR.pdf - Published Version Restricted to Registered users only Download (16MB) | Request a copy

Abstract

A time series is a sequence of observations that are time-oriented or arranged chronologically for an observed variable. Time series data are widely used in analysis and forecasting, including stock price analysis. Stock price data are generally non-stationary, fluctuating, highly volatile, and tend to form nonlinear patterns. These characteristics cause the assumptions of classical time series methods, particularly Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU), which are used to predict stock prices using univariate and multivariate approaches. The data used in this study were the stock price data of PT Bank Mandiri Tbk (BMRI) from 2015 to 2024. The research stages included Exploratory Data Analysis (EDA), preprocessing, modeling, and architecture optimization, with evaluation conducted using Mean Squared Error (MSE) and Mean Absolute Error (MAE). The results showed that both LSTM and GRU models with univariate and multivariate approaches were able to predict stock prices with movements relatively similar to the actual data. The univariate GRU model was identified as the best architecture consisting of a sliding window of 30, one hidden layer, 128 units, dropout of 0.1, batch size of 32, and MSE and MAE values of 0.0012 and 0.0270, respectively. In addition, GRU demonstrated higher sensitivity than LSTM, resulting in more aggressive predictions of price declines, while LSTM produced predictions that were more realistic.

Share this knowledge with your friends :

Actions (login required)

View Item