Reza Taleblou
Abstract
In this paper, we evaluate the performance of two machine learning architectures— Recurrent Neural Networks (RNN) and Transformer-based models—on four commodity-based company indices from the Tehran Stock Exchange. The Transformer-based models used in this study include AutoFormer, FEDformer, ...
Read More
In this paper, we evaluate the performance of two machine learning architectures— Recurrent Neural Networks (RNN) and Transformer-based models—on four commodity-based company indices from the Tehran Stock Exchange. The Transformer-based models used in this study include AutoFormer, FEDformer, Informer, and PatchTST, while the RNN-based models consist of GRU and LSTM. The dataset comprises daily observations collected from April 20, 2020, to November 20, 2024. To enhance the generalization power of the models and prevent overfitting, we employ two techniques: splitting the training and test samples, and applying regularization methods such as dropout. Hyperparameters for all models were selected using a visual method. Our results indicate that the PatchTST model outperforms other methods in terms of Root Mean Squared Error (RMSE) for both 1-day and 5-day (1-week) forecasting horizons. The FEDformer model also demonstrates promising performance, particularly for forecasting the MetalOre time series. In contrast, the AutoFormer model performs relatively poorly for longer forecasting horizons, while the GRU and LSTM models yield mixed results. These findings underscore the significant impact of model selection and forecasting horizon on the accuracy of time series forecasts, emphasizing the importance of careful model choice and hyperparameter tuning for achieving optimal performance.
Roya Karimkhani; Yousef Edrisi Tabriz; Ghasem Ahmadi
Abstract
Forecasting price trends in financial markets is of particular importance for traders because price trends are inherently dynamic and forecasting these trends is complicated. In this study, we present a new hybrid method based on combination of the dynamic mode decomposition method ...
Read More
Forecasting price trends in financial markets is of particular importance for traders because price trends are inherently dynamic and forecasting these trends is complicated. In this study, we present a new hybrid method based on combination of the dynamic mode decomposition method and long short-term memory method for forecasting financial markets. This new method is in this way that we first extract the dominant and coherent data using the dynamic mode decomposition method and then predict financial market trends with the help of these data and the long short-term memory method. To demonstrate the efficacy of this method, we present three practical examples: closing price of US Dollar to Iranian Rial, closing prices of zob roy Isfahan stock, and also closing prices of siman shargh stock. These examples exhibit bullish, bearish, and neutral behaviors, respectively. It seems that the proposed new method works better in predicting the financial market than the existing long-short-term memory method.
Kamran Pakizeh; Arman Malek; Mahya Karimzadeh khosroshahi; Hasan Hamidi Razi
Abstract
Cryptocurrencies, which are digitally encrypted and decentralized, continue to attract attention of nancial market players across the world. Because of high volatility in cryptocurrency market, predicting price of cryptocurrencies has become one of the most complicated elds in nancial ...
Read More
Cryptocurrencies, which are digitally encrypted and decentralized, continue to attract attention of nancial market players across the world. Because of high volatility in cryptocurrency market, predicting price of cryptocurrencies has become one of the most complicated elds in nancial markets. In this paper, we use Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) models to predict price of four well-known cryptocurrencies of Bitcoin (BTC), Ethereum(ETH), Litecoin (LTC), and Ripple (XRP). These models are subdivisions of Articial Intelligence, machine learning and data science. The main aim of this paper is to compare the accuracy of above-mentioned models in forecasting time series data, to nd out which model can better predict price in these four cryptocurrencies. 43 variables consisting of 28 technical indicators and t+10 lags were calculated and appended to the Open, High, Low, Close and Volume (OHLCV) data for selected cryptocurrencies. Applying random forest as feature selection, 25 variables werechosen, 24 of them selected as feature (independent variables) and one as a dependent variable. Each attribute value was converted into a relative standard score, followed by Min-max scaling; we compare models and results of Dieblod Mariano test that is used to examine whether the differences in predictive accuracy with these two models are signi cant, reveal that LSTM reaches better accuracy than GRU for BTC and ETH, but both models convey the same accuracy for LTC and XRP.
