Performance Comparison Of Gated Recurrent Unit (Gru) And Gated Recurrent Hybrid Models Unit (Gru) On Bri Stock Forecasting (Case Study: Pt. Bank Rakyat Indonesia Tbk (Bbri) Share Price Period January 2, 2019-october 18, 2023)

Abstract

Indonesia, as a developing country rich in resources, demonstrates positive economic growth with the banking sector playing a crucial role in driving economic advancement. This study focuses on PT. Bank Rakyat Indonesia Tbk (BBRI). The objective of this research is to evaluate the effectiveness of using decomposition first by retaining the seasonal index or solely utilizing the Gated Recurrent Unit (GRU). The issues addressed include the description of BBRI stock data, the forecasting process using the hybrid decomposition-GRU method and GRU, and determining the best method for forecasting BBRI stock data. This study depicts efforts to understand and predict stock price trends by comparing two different approaches using BBRI stock price data with a sample size of 1,169. The use of hyperparameters such as 32 hidden units and a batch size of 512 with the Relu activation function in the hybrid decomposition-GRU model architecture shows better accuracy compared to the GRU model architecture, as the hybrid decomposition-GRU model yields the smallest MAPE value of 6.824051%.