Hirad Dolatzadeh

Hirad Dolatzadeh

AI/ML Professional | 3+ years ML Engineering | Generative AI Specialist | MBA (4.0 GPA) | Open to Relocation

Bitcoin Return Prediction Using Sentiment Analysis

How People's Sentiment and Attention Affect the Return of Bitcoin?

Oct 2021 – Sep 2022 | Sharif University of Technology | GitHub Repository

Research Overview

This MBA capstone project investigated the relationship between investor sentiment, market attention, and Bitcoin returns using advanced machine learning and NLP techniques.

Key Findings:

  • Predictive Accuracy: 62% - Developed models capable of forecasting Bitcoin returns with 62% accuracy
  • Identified key variables including the Fear and Greed Index, Twitter sentiment, and Google search trends
  • Demonstrated the significant impact of investor behavior on cryptocurrency market dynamics

Methodologies:

  • NLP Techniques: Twitter sentiment analysis, text mining
  • Deep Learning Models: TensorFlow, Keras, RNNs
  • Explainability: SHAP (SHapley Additive exPlanations) for model interpretability
  • Data Sources: Google Trends, Twitter data, Fear and Greed Index

Impact:

This research contributed to understanding the behavioral finance aspects of cryptocurrency markets and demonstrated how alternative data sources can enhance financial prediction models.