Cryptocurrency Analysis using Social Media Sentiments

Cryptocurrency Analysis using Social Media Sentiments

Nishchal Nandagopal , Atla Sai Abhinav , Karthik Sai BS , Kumar Ketan

Computer Science and Engineering, BMS College of Engineering,Bengaluru, Karnataka-560019

Abstract

The  rapid  emergence  and  widespread  adoption  of  digital  media,  particularly  social  media  platforms,  have  sparked  a  profound revolution in technology. This transformation has not only revolutionized the way we communicate and share information but has also opened up new avenues for understanding human emotions, sentiments, and opinions. In this dynamic landscape, the present study aims to delve deeper into the intricate fluctuations observed in Bitcoin prices within a specific timeframe and subsequently harness the power of machine learning to predict future values.  To achieve this ambitious objective, the study capitalizes on the vast trove of user comments available on Reddit, one of the most influential social media platforms of our time.

The study adopts a cutting-edge deep learning framework that leverages the prowess of a Bidirectional Recurrent Neural Network (RNN) combined with the robustness of Long Short-Term Memory (LSTM) units.  Furthermore, to enhance the model’s capacity to comprehend and represent the complex semantic structures inherent in textual data, word2Vec embeddings are employed.  This powerful combination of advanced neural network architectures and sophisticated word embeddings sets the stage for accurate and insightful predictions.

To ensure the reliability and efficacy of the predictions, the textual data obtained from Reddit undergoes a meticulous refinement process, carefully removing noise and extracting the most relevant and meaningful information. This refined data is then harnessed to train the model, enabling it to make accurate forecasts of Bitcoin prices across various time intervals.  The model’s predictive capabilities extend to time horizons of +1,  +2,  +6,  +12,  and +24 hours,  providing valuable insights into the short-term price dynamics of Bitcoin.

The true power of the model lies in its ability to directly forecast the direction of price change, allowing it to uncover hidden patterns and underlying trends in the price movement.   By deciphering the intricate relationships between user comments and Bitcoin price fluctuations, the model gains the expertise to make informed predictions that capture the essence of market dynamics. In summary, the advent of digital media, coupled with advancements in deep learning techniques, has paved the way for ground- breaking research into the realm of cryptocurrency prediction.  By harnessing the wealth of user comments on social media plat- forms and employing advanced neural network architectures, this study exemplifies the potential to unravel the complex dynamics

of Bitcoin prices and offer valuable insights into future market trends.