JUPYTER NOTEBOOKS

Ready-to-use notebooks for common memecoin analysis tasks. Download, run locally, and adapt to your needs.

Getting Started

1. Download the free sample dataset (300 tokens)

2. Install required dependencies:

pip install pandas numpy scikit-learn xgboost duckdb matplotlib seaborn

3. Download a notebook and open in Jupyter:

jupyter notebook rug_detection_model.ipynb

💡 All notebooks are designed to work with the free sample dataset. For production models, request full access.

Rug Detection Model

Train a machine learning model to predict rug pulls using the 54 confirmed rugs in the sample dataset

Intermediate•30 min

What You'll Learn

▸Load and explore the BSC memecoin dataset
▸Feature selection from 129 ML parameters
▸Train XGBoost classifier on rug/non-rug labels
▸Evaluate model performance (precision, recall, F1)
▸Feature importance analysis
▸Predict rug probability for new tokens

Key Metrics & Insights

•54 rugs vs 246 non-rugs in training set
•Top features: gini_coefficient, contract_suspicious_score, early_buy_count
•Expected accuracy: ~85% on validation set

ClassificationXGBoostFeature Engineering

Sniper Wallet Identification

Identify and analyze high-performance sniper wallets based on early buy patterns and profitability

Beginner•20 min

What You'll Learn

▸Query wallet_actions table for early buyers
▸Calculate wallet-level profitability metrics
▸Identify wallets with consistent early entries
▸Analyze correlation between entry timing and returns
▸Visualize sniper wallet behavior patterns
▸Export high-reputation wallet addresses

Key Metrics & Insights

•Analyze early_buy_count and early_buy_unique_wallets
•Track wallets across multiple token launches
•Identify top 1% performers by max_return_24h

AnalysisPandasVisualization

Optimal Entry Timing Prediction

Predict the best entry timing for memecoin trades using holder distribution snapshots and transaction patterns

Advanced•45 min

What You'll Learn

▸Load holder distribution time series (30s, 1min, 5min snapshots)
▸Engineer temporal features from transaction data
▸Build LSTM model for entry timing prediction
▸Backtesting framework for strategy evaluation
▸Risk-adjusted return optimization
▸Real-time inference pipeline design

Key Metrics & Insights

•Use holder_count_30s, holder_count_1min, holder_count_5min
•Optimize for max_return_24h while minimizing drawdown
•Target: 2x improvement over random entry timing

Time SeriesLSTMDeep Learning

Ready to Build Your Own Models?

These notebooks are just the beginning. With full dataset access, you can train production-grade models on thousands of tokens with complete transaction history.