كوثر نصري

Bitcoin price prediction, as one of the most important and volatile cryptocurrencies in the world, has always been one of the most complex and fundamental challenges in financial market analysis. This complexity primarily arises from the inherent instability of the cryptocurrency market and its susceptibility to a wide range of economic, political, social, and technological factors. Accurate and reliable modeling for predicting the price of this cryptocurrency requires the use of advanced, data-driven analytical approaches. In response to this need, this research presents a precise model for Bitcoin price prediction, a hybrid model based on deep neural networks. This model, leveraging Bitcoin's price and transactional data, along with valuable insights from the blockchain, is capable of identifying and analyzing nonlinear relationships and hidden patterns within the data. The proposed model architecture includes the use of a bidirectional Long Short-Term Memory (LSTM) network combined with an attention mechanism, a combination that has proven effective in extracting time-dependent features and focusing on more important data segments in deep learning literature. The data used in this research was collected on a daily basis, and after preprocessing and normalization, it was used as input to the model. The results from experimental tests demonstrate the remarkable performance of the proposed model in predicting Bitcoin's price, with eva‎luation metrics including Root Mean Squared Error (RMSE) and Mean Absolute Percentage Error (MAPE) being 52.116 and 95.1, respectively. These results indicate that the developed model has a significant relative superiority in predictive accuracy compared to traditional methods and baseline models.

داده‌هاي درون زنجيره‌اي , زنجبره بلوكي , پيش‌بيني قيمت , رمزارزها , مدل‌هاي يادگيري ماشين

On_Chain dats , Blockchain , Price Prediction , cryptocurrency , machine learning model

Kowsar Nasri

DR Reza Entezari Maleki