Abstract:
"Predicting visual prices remains a challenging task due to its highly volatile nature, prompting exploration of advanced techniques such as Feedforward Neural Networks (FNNs). Sentiment analysis and FNN models along with XAI co-utilization for Bitcoin price prediction is yet a relatively unexplored field.. Sentiment analysis of social media data, particularly from platforms like Twitter, offers valuable insights into market psychology. However, the interplay between sentiment and Bitcoin price remains a relatively under-investigated area. This study also aims to address this gap by analyzing the correlation between sentiment scores derived from Twitter and historical Bitcoin price data. To achieve this, Author collects sentiment ratings from Twitter alongside historical Bitcoin price data. A preprocessing stage involving tokenization, stop word removal, and elimination of rows without tweets ensures data cleanliness. Subsequently, timestamps are matched, and the datasets are merged to align sentiment scores with corresponding Bitcoin prices. This allows to analyze the correlation between these two variables and showcase how social media sentiment might affect Bitcoin price movement. eXplainable Artificial Intelligence (XAI)—more specifically, the Local Interpretable Model- agnostic Explanations (LIME) framework—further improves this approach by offering transparent insights into the prediction process. The predictive performance of the model is evaluated quantitatively using metrics like Mean Squared Error (MSE) and Root Mean Squared Error (RMSE), Mean Absolute error(MAE), R2 Score and Variance Regression score offer information on what level of accuracy is the predicted outcomes.
In conclusion, this study offers a novel approach to enhance Bitcoin price prediction models by combining FNNs with XAI approaches, along with showcasing the impact of social media sentiment on the bitcoin price."
