Forecasting the Stock Price of Bank Central Asia Using Singular Spectrum Analysis Based Model

This paper applied a Singular Spectrum Analysis (SSA) and a hybrid model of SSA and Neural Network (NN) to forecast stock prices of BCA. SSA is regarded as effective in capturing the deterministic components of the data, whereas NN is used to model the stochastic elements. The data that used are stock prices of BCA from June 2023 to February 21, 2024. The data are forecasted using SSA-LRF and SSA-NN hybrid models. The hybrid model is performed in two ways. First, the data are modeled using SSA, and the residual of the SSA-LRF is modeled using NN. Second, the results of SSA decomposition, namely the signal and noise components, are modeled using NN. The results show that the best model is the SSA-NN hybrid model with a window length of 79 and network architecture of (1,64,1) for the signal component and (50,64,1) for the noise component, and can reduce the MAPE for the testing data from that obtained by SSA-LRF up to 36%.