Predicting Capital Market and Cryptocurrency Performance in Higher Sequences Using Deep Learning

Abstract The prediction of price trend fluctuations in financial markets has consistently attracted the attention of researchers and financial market participants. Accordingly, the present study aims to design a price trend prediction model using a deep learning technique to estimate the next seven sequences namely, the closing prices over the next seven days for both Bitcoin as a representative of the cryptocurrency market and the Tehran Stock Exchange All-Share Index (TEDPIX) as an indicator of the Iranian capital market. This research is applied in terms of purpose, descriptive-correlational in terms of methodology, and based on time series data. To achieve the research objectives, data from 1,019 consecutive days of Bitcoin closing prices and the TEDPIX, up to 20 December 2024, were utilized. For data analysis, a deep learning technique, one of the major branches of artificial intelligence, was implemented within the Python programming environment. Based on the research findings, it can be stated that overall, as the model advances to higher-order sequences, the estimation error in predicting the TEDPIX increases slightly. However, given the marginal nature of the error increase, the designed predictor remains reliable for future sequence forecasting of this index. Conversely, in the case of Bitcoin, the estimation error shows a steeper rise across sequential predictions compared to the Iranian capital market. This indicates that, based on this predictor, the investment risk in higher-order sequences is greater for Bitcoin.