An empirical cryptocurrency price forecasting model

The goal of this manuscript is to use deep learning based multi-modal hybrid model to forecast the value of bitcoins by analyzing the influence of social media attitudes. In order to analyze, roughly two thousand samples of on-chain, cryptocurrency market, and related social media information (Twitter/X) are gathered between the years 2014 and 2022. The attitudes shared on social media regarding bitcoin are then analyzed utilizing the Twitter-RoBERTa (robustly optimized bidirectional encoder representation from transformers) and VADER (value-aware dictionary for sentiment reasoning) algorithms after we incorporated sentiment data from Twitter into the models. The proposed model produces the least MAPE (4.37%) and least RMSE (6.55%) thereby demonstrating its ability to accurately forecast market trends with least fluctuations from the available dataset. By illustrating the promise of social networking sentiment evaluation, on-chain data collection, and the use of long short-term memory (LSTM) neural network for forecasting market patterns, this study makes an important addition to the field of economic forecasting and offers traders, agents, and investors a useful tool for making informed choices.