Skip to content

QUICKSTART.md

Latest commit

 

History

History
130 lines (98 loc) · 2.86 KB

QUICKSTART.md

File metadata and controls

130 lines (98 loc) · 2.86 KB

Quick Start Guide

Launch the UI (Recommended)

# From project directory
./run_ui.sh

# Or manually
streamlit run app.py

Opens at: http://localhost:8501

First Run Workflow

  1. Start UI → Browser opens automatically
  2. Go to Dashboard → See current state (empty if new)
  3. Configure & Run → Single Run tab
  4. Adjust parameters → Use defaults or experiment
  5. Click "Run Single Configuration" → Wait for completion
  6. Results Explorer → View metrics and details
  7. Visualizations → See cluster patterns

Quick Commands

Run from CLI

python main.py

Install Dependencies

pip install -r requirements.txt

View Results

# All metrics
cat results/metrics/metrics.csv

# Latest log
tail -f logs/*.log | tail -1

Generate Visualizations

# From Python
from tools.visualize_clusters import plot_cluster_samples
plot_cluster_samples(run_id=1, cluster_ids=[0,1], n_samples=5)

Key Directories

  • data/ - Your OHLCV CSV files
  • results/metrics/ - Metrics and summaries
  • results/models/ - Saved HDBSCAN models
  • results/labels/ - Cluster assignments
  • results/visualizations/ - Generated plots
  • logs/ - Execution logs

UI Pages

Page Purpose
🏠 Dashboard Overview and trends
🛠 Configure & Run Start experiments
📊 Results Explorer Analyze runs
📈 Visualizations View patterns
📄 Logs Debug and monitor
💾 Model Manager Manage storage

Common Parameters

Parameter Typical Range Description
Window Size 10-15 Bars per pattern
Min Cluster Size 10-15 Minimum samples in cluster
Min Samples 5-10 Core point neighbors
Metric euclidean Distance measure

Interpreting Results

Good Run:

  • Silhouette Score > 0.6
  • Noise Ratio < 10%
  • Multiple distinct clusters (2-10)

Needs Tuning:

  • Silhouette Score < 0.4
  • Noise Ratio > 20%
  • Only 1 cluster or all noise

Troubleshooting

UI won't start:

pip install streamlit plotly
streamlit run app.py --server.port 8502

No GPU detected (optional):

pip install cupy-cuda12x cuml-cu12

Out of memory:

  • Reduce dataset size in Configure & Run
  • Delete old models in Model Manager
  • Uncheck "Save Features" option

Next Steps

  1. Read docs/UI_GUIDE.md for detailed features
  2. Check docs/DESIGN.md for architecture
  3. Experiment with grid search for parameter tuning
  4. Upload your own OHLCV data

Support Files

  • README.md - Full project overview
  • docs/UI_GUIDE.md - Complete UI documentation
  • docs/DESIGN.md - Technical architecture
  • requirements.txt - Python dependencies

You're ready to go! 🚀

Launch the UI with ./run_ui.sh and start exploring your OHLCV patterns.