|
|
 |
| |
|
|
|
A Comparative Statistical Study of Time Series and Deep Learning Models for Short-Term Stock Price Forecasting |
|
|
|
PP: 803-815 |
|
|
doi:10.18576/jsap/150330
|
|
|
|
Author(s) |
|
|
|
Sarah Aljohani,
Amir Ahmad Dar,
I. Silambarasan,
|
|
|
|
Abstract |
|
|
| In this study, we investigate the performance of different predictive models in forecasting the closing stock price
of RTPower Ltd. over a 15-day horizon, a task with significant implications for investment policy and risk management given the centrality of financial prediction to modelling practice. Historical data on daily stock prices from the National Stock Exchange was gathered and pre-processed, and training and testing of different models were carried out, including time- series and machine learning algorithms. The models were evaluated based on standard evaluation metrics such as MAE, MSE, RMSE, and MAPE. It was shown that LSTM was more successful in comparison with other models because it showed the lowest percentage of errors and the highest degree of predictive success. LSTM achieved the lowest errors across all metrics (MAE: 1.4621, RMSE: 1.7458, MAPE: 11.17%), outperforming SARIMA (MAPE: 25.64%), XGBoost (MAPE: 14.71%), Prophet (MAPE: 38.37%), and Transformer (MAPE: 43.27%). This is possibly due to its capability to learn temporal dependencies in the data. The chosen model was applied to determine the closing prices for the following 15 days, and it proved highly practical for short-term trading strategies. This paper also highlights the significance of higher-level modelling tools in predicting financial time series and, simultaneously, underscores the necessity of selecting a model suited to the specific prediction problem. The constraints involving susceptibility to market fluctuations and data quality are addressed, along with suggestions for future studies, including the incorporation of external variables such as market
sentiment. The findings contribute to the existing literature on stock price prediction and offer practical guidance to investors. |
|
|
|
|
|
|
|
