Resumen

GARCIA, Samuel. Comparing the Performance of Long Short-Term Memory Architectures (LSTM) in Equity Price Forecasting: A Research on the Mexican Stock Market. The Anáhuac j. [online]. 2024, vol.24, n.1, pp.160-179.  Epub 26-Ago-2024. ISSN 2683-2690.  https://doi.org/10.36105/theanahuacjour.2024v24n1.06.

This study compares the performance of univariate and multivariate Long Short-Term Memory (LSTM) to predict next-day closing prices on four stocks in the consumer retail sector of the Mexican Stock Exchange. Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), Median Absolute Percentage Error (MdAPE), and Root Mean Squared Error (RMSE) are used to test the networks’ performance. Results show a better performance on multivariate price forecasts when using 20-day and 15-day length sequences, generating consistent results for the sample, including illiquid and liquid stocks. On the other hand, univariate LSTM discloses lower forecast performance when predicting the price of illiquid stocks.

Palabras llave : forecast; stocks; univariate; multivariate; LSTM.

        · resumen en Español     · texto en Inglés     · Inglés ( pdf )