International Conference on Information and Knowledge Technology

فارسی

Home / شانزدهمین کنفرانس بین المللی فناوری اطلاعات و دانش

CRYPTOCURRENCY PRICE PREDICTION USING A HYBRID DEEP MODEL AND TECHNICAL AND PSYCHOLOGICAL INDICATORS

Authors :

Mohammadreza Borjian¹ Mohammad Mehdi Homayounpour²

1- موسسه آموزش عالی شهاب دانش قم 2- موسسه آموزش عالی شهاب دانش قم

Keywords :

Cryptocurrency،Deep Learning،Time Series Forecasting،Transformer

Abstract :

This study develops a hybrid model for daily price prediction based on key and influential features of the financial market. The proposed model integrates Transformer architecture, Convolutional Neural Networks (CNN), and Gated Recurrent Unit (GRU) networks to provide improved performance compared to previous research in forecasting the next day's prices. The model combines the CNN's capability in extracting essential features from time series data and GRU's ability to retain temporal dependencies. The Transformer structure enables the model to identify complex spatio-temporal features more effectively, while the GRU maintains long-term dependencies among data. In this study, the Fear and Greed Index—a measure of market sentiment—is incorporated as a key input into the proposed hybrid model. By combining this psychological indicator with traditional market data such as price and trading volume, the model gains the ability to capture not only historical price trends but also the emotional and behavioral dynamics influencing market movements.A nine-day time-series dataset including previous prices, the Fear and Greed Index, the Relative Strength Index (RSI), and trading volume was employed for training and evaluation. The results demonstrate that the proposed model achieves higher accuracy in price prediction compared to previous models and can detect more complex and nonlinear patterns in financial data. This model can serve as an efficient tool for financial analysts and investors to make more informed and effective investment decisions in financial markets.