📊 Crypto-Sentiment-Trading-Analysis

This project explores the relationship between Bitcoin market sentiment (based on the Fear & Greed Index) and real-world trader performance from the Hyperliquid trading platform. It includes data analysis, visualizations, and a machine learning model to predict trade profitability based on market sentiment and trade-related features.

---

🔍 Problem Statement

Can market sentiment indicators (like "Fear" and "Greed") help predict whether a crypto trade will be profitable?

---

📁 Dataset Information

🗂 1. Fear & Greed Index

• Columns: date, value, sentiment
• Simulated sentiment was used because original data did not cover the recent time window of trader data.

🗂 2. Hyperliquid Trader Data

• Columns: Account, Symbol, Execution Price, Size Tokens, Side, Timestamp, Closed PnL, etc.
• Timestamp in milliseconds was parsed to generate date.

---

🔧 Tools Used

• Python
• Pandas, NumPy
• Matplotlib, Seaborn
• Scikit-learn (for ML modeling)

---

📊 Exploratory Data Analysis

• Bar plot: Average Closed PnL grouped by sentiment (Fear, Greed, etc.)
• Buy ratio distribution by sentiment group
• Correlation matrix between sentiment and Closed PnL

---

🧠 Machine Learning Model

• Model: RandomForestClassifier
• Features used:
  • sentiment_score (encoded from sentiment)
  • Execution Price
  • Size Tokens
  • Trade Side (buy/sell encoded)
• Class balancing was performed to fix the model bias toward loss-making trades.

🔎 Results:

• Accuracy: 77%
• F1-Score: 0.77
• Both profit and loss trades are now correctly classified.

---

📌 Key Insights

• Market sentiment alone shows very weak correlation with profit (r ≈ 0.004).
• Adding trade-specific features improves prediction significantly.
• Traders show slightly higher profit during periods of "Greed".
• Buy vs. sell ratios remain fairly balanced across sentiment groups.