ManekiNeko（招财猫）

this project is used to explore the possibility of lying to earn money in crypto currency

English | 中文

including following parts:

• getTradeData get trade data from binance and save to sqlite
• lstmTrain using lstm to predict the price of crypto currency
• xGBoost using XGBoost to predict the price of crypto currency
• randomForest using random forest to predict the price of crypto currency
• transformer using Transformer to predict the price of crypto currency
• informer using Informer to predict the price of crypto currency
• LightGBM / CatBoost efficient gradient boosting variants
• N-BEATS deep neural architecture for interpretable time series forecasting
• Reinforcement Learning (PPO/DQN) training agents to make trading decisions directly

Implementation Roadmap: Detailed plan for implementing the above advanced models.

Feature Engineering & Signal Generation

Technical Indicators

The project generates extensive technical indicators to capture market dynamics:

• Momentum: RSI (14-period)
• Trend: EMA (12/26), MACD (Signal, Histogram), SMA (3, 6, 12, 20)
• Volatility: Bollinger Bands (Upper, Middle, Lower), Volatility Std (5-period)
• Price Differentials: Distance from SMA and Bollinger Bands

Signal Labeling Logic

The model treats trading as a multi-class classification problem with signals derived from local extrema detection:

1. Fixed Interval Extrema (Primary Signals)

Data is divided into non-overlapping windows (default size: 300).

• Label 4 (Buy): Local Minimum (Entry point)
• Label 6 (Sell): Local Maximum (Exit point) Note: The algorithm pairs minima with subsequent maxima to ensure valid trade cycles.

2. Sliding Window Extrema ("Slip" Signals)

A rolling window scans the time series to capture more granular market turning points.

• Label 2 (Slip Buy): Local Minimum in sliding window
• Label 3 (Slip Sell): Local Maximum in sliding window

3. Default State

• Label 1 (Hold): No specific signal detected

getTradeData usage

copy config_example.ini to config.ini and change the api key and secret key

pip install requirements.txt
python getTradeData.py

license

MIT