Exportar Publicação

Mendes, D. A., Ferreira, N. B. & Mendes, V. (2023). Could data frequency imply better forecast performance for stock markets? A case study for G7 economies. 9th International conference on Time Series and Forecasting.

D. E. Mendes et al.,  "Could data frequency imply better forecast performance for stock markets? A case study for G7 economies", in 9th Int. conference on Time Series and Forecasting, Las Palmas, 2023

@misc{mendes2023_1732361246002,
	author = "Mendes, D. A. and Ferreira, N. B. and Mendes, V.",
	title = "Could data frequency imply better forecast performance for stock markets? A case study for G7 economies",
	year = "2023",
	howpublished = "Digital",
	url = "https://itise.ugr.es/"
}

TY  - CPAPER
TI  - Could data frequency imply better forecast performance for stock markets? A case study for G7 economies
T2  - 9th International conference on Time Series and Forecasting
AU  - Mendes, D. A.
AU  - Ferreira, N. B.
AU  - Mendes, V.
PY  - 2023
CY  - Las Palmas
UR  - https://itise.ugr.es/
AB  - The analysis and prediction of stock markets are critical issues for suitable investments and financial cooperation. However, due to non-stationary, high volatility, and complex nonlinear patterns of stock market fluctuation, it is pretty demanding to predict the stock price accurately. Nowadays, hybrid and ensemble models based on machine learning and economics replicate several patterns learned from the time series and sum the relevant information, improving the forecast performance. 
This paper discusses and analyses different models for stock price forecasting. First, we use SARIMAX models in a classical approach and by using AutoML algorithms from the Darts library. Second, a deep learning procedure predicts the stock prices for the seven world's most representative economies (G7). In particular, LSTM (Long Short Term Memory) and BiLSTM recurrent neural networks (with and without stacking), with optimized hyperparameters architecture by KerasTuner, in the context of different time-frequency data (with and without mixed frequencies) are implemented.
Several research papers and reports deal with daily data. However, this issue of daily data can be largely dependent on deterministic variables like day-of-week, week-of-the-year, month-of-the-year, week-of-the-month, and long weekends. Furthermore, there may be changes in daily patterns and different volatilities for additional days of the week due to macroeconomic factors, fundamental factors, and investor sentiment. Consequently, it follows the high interest in the multi-step-ahead stock price index forecasting by using different time frequencies (daily, one-minute, five-minute, and ten-minute granularity), focusing on raising intraday stock market prices.  
The results show that the BiLSTM model forecast outperforms the benchmark models –the random walk and SARIMAX - and slightly improves LSTM. More specifically, the average reduction in error rates by BiLSTM is 14-17 percent compared to SARIMAX. According to the scientific literature, we also obtained that high-frequency data improve the forecast accuracy by 3-4% compared with daily data since we have some insights about the volatility driving forces.

ER  -