Stock Price Prediction through STL Decomposition using Multivariate Two-way Long Short-term Memory

With advancements in machine-learning techniques, stock-price movements can ostensibly be forecasted using time-series data. In this study, several different types of long short-term memory (LSTM) are used to predict the closing prices Japanese stocks five days into future. Also, four features [i.e., simple moving average (SMA), linear weighted (WMA), exponential WMA (EMA), and Savitzky–Golay (SG) metric] generated from daily data split two components (i.e., trend seasonal) by applying seasonal–trend decomposition Loess (STL) decomposition. The prediction results evaluated terms return, root-mean-square error (RMSE), mean absolute (MAE), other relevant measures accuracy relevancy. As a result, multivariate two-way LSTM model yielded highest overall performance. respect RMSE MAE training data, was not superior models. However, with on validation it best. performance direction stock prices.