Skip to content

Latest commit

 

History

History
74 lines (51 loc) · 1.73 KB

QUICK_START.md

File metadata and controls

74 lines (51 loc) · 1.73 KB

Quick Start Guide

🚀 Get Started in 5 Minutes

Step 1: Install

npm install
npm run build

Step 2: Verify Data

Make sure merged_dataset.csv is in the project root.

Step 3: Train

# Best option for crypto (recommended)
npm run train -- --enhanced --returns

# Or try standard training
npm run train

Step 4: View Results

  • Check console for metrics
  • Open validation_results.html in browser

📊 What to Expect

Good Results:

  • MAPE: 20-30%
  • R²: 0.3-0.5
  • Predictions follow general trend

Acceptable Results:

  • MAPE: 30-40%
  • R²: 0.1-0.3
  • Some prediction errors

Note: 30% error is normal for crypto next-day prediction due to high volatility.

🎯 Training Options

Command Description Best For
npm run train Standard ensemble Baseline comparison
npm run train -- --enhanced TEMPO with decomposition Feature enrichment
npm run train -- --enhanced --returns Predicts returns Volatile crypto (recommended)

📖 Full Guide

For detailed instructions, see USER_GUIDE.md

❓ Troubleshooting

Problem: "merged_dataset.csv not found"

  • Fix: Make sure CSV file is in project root

Problem: High error (40%+)

  • Fix: Try --returns flag: npm run train -- --enhanced --returns

Problem: Training too slow

  • Fix: Reduce rollingWindow in src/scripts/train.ts

💡 Tips

  1. Start with return-based training - usually best for crypto
  2. Compare methods - try standard vs enhanced
  3. Check HTML results - visual charts help understand performance
  4. Be patient - training takes 10-30 minutes

Ready? Run: npm run train -- --enhanced --returns