オープンソースライブラリ
Freqtrade
freqtrade/freqtrade★ 28k+
Crypto trading bot framework with FreqAI ML extension supporting multiple machine learning and deep learning models.
インジケーターベースのテンプレート
機械学習モデル（FreqAI）
機械学習ベースの価格予測のための FreqAI 予測モデル。各モデルは Freqtrade の特徴量パイプラインを通じた特徴量エンジニアリングが必要です。
Gradient boosting regression model for price movement prediction using LightGBM.
速度⚡⚡⚡
精度📊📊📊
主要パラメータ
| n_estimators | 1000 | Number of boosting rounds |
| learning_rate | 0.01 | Step size shrinkage |
ソース：
freqai/prediction_models/LightGBMRegressor.pyGradient boosting classification model for directional prediction (up/down/neutral).
速度⚡⚡⚡
精度📊📊📊
主要パラメータ
| n_estimators | 1000 | Number of boosting rounds |
ソース：
freqai/prediction_models/LightGBMClassifier.pyXGBoost-based regression model for continuous value prediction.
速度⚡⚡⚡
精度📊📊📊
主要パラメータ
| n_estimators | 1000 | Number of boosting rounds |
| max_depth | 6 | Maximum tree depth |
ソース：
freqai/prediction_models/XGBoostRegressor.pyXGBoost-based classification model for directional prediction.
速度⚡⚡⚡
精度📊📊📊
主要パラメータ
| n_estimators | 1000 | Number of boosting rounds |
ソース：
freqai/prediction_models/XGBoostClassifier.pyCatBoost gradient boosting model with native categorical feature support.
速度⚡⚡
精度📊📊📊
主要パラメータ
| iterations | 1000 | Number of boosting iterations |
ソース：
freqai/prediction_models/CatboostRegressor.pyMulti-layer perceptron neural network for regression-based price prediction.
速度⚡⚡
精度📊📊📊
主要パラメータ
| hidden_dim | 128 | Hidden layer dimension |
| dropout_percent | 0.2 | Dropout rate |
ソース：
freqai/prediction_models/PyTorchMLPRegressor.py強化学習（FreqAI）
Stable Baselines3 上に構築された強化学習エージェント。ラベル付きデータではなく環境との相互作用を通じてエージェントが取引判断を学習します。
Freqtrade 戦略と FreqAI モデル, アルゴリズムリファレンス
- SampleStrategy (Freqtrade)
- Template strategy demonstrating basic technical indicator usage for entry/exit signals. ソース: https://github.com/freqtrade/freqtrade/blob/develop/freqtrade/templates/sample_strategy.py.
- LightGBMRegressor (Freqtrade)
- Gradient boosting regression model for price movement prediction using LightGBM. 主要パラメータ: n_estimators (Number of boosting rounds), learning_rate (Step size shrinkage).ソース: https://github.com/freqtrade/freqtrade/blob/develop/freqai/prediction_models/LightGBMRegressor.py.
- LightGBMClassifier (Freqtrade)
- Gradient boosting classification model for directional prediction (up/down/neutral). 主要パラメータ: n_estimators (Number of boosting rounds).ソース: https://github.com/freqtrade/freqtrade/blob/develop/freqai/prediction_models/LightGBMClassifier.py.
- XGBoostRegressor (Freqtrade)
- XGBoost-based regression model for continuous value prediction. 主要パラメータ: n_estimators (Number of boosting rounds), max_depth (Maximum tree depth).ソース: https://github.com/freqtrade/freqtrade/blob/develop/freqai/prediction_models/XGBoostRegressor.py.
- XGBoostClassifier (Freqtrade)
- XGBoost-based classification model for directional prediction. 主要パラメータ: n_estimators (Number of boosting rounds).ソース: https://github.com/freqtrade/freqtrade/blob/develop/freqai/prediction_models/XGBoostClassifier.py.
- CatboostRegressor (Freqtrade)
- CatBoost gradient boosting model with native categorical feature support. 主要パラメータ: iterations (Number of boosting iterations).ソース: https://github.com/freqtrade/freqtrade/blob/develop/freqai/prediction_models/CatboostRegressor.py.
- PyTorchMLPRegressor (Freqtrade)
- Multi-layer perceptron neural network for regression-based price prediction. 主要パラメータ: hidden_dim (Hidden layer dimension), dropout_percent (Dropout rate).ソース: https://github.com/freqtrade/freqtrade/blob/develop/freqai/prediction_models/PyTorchMLPRegressor.py.
- PyTorchTransformerRegressor (Freqtrade)
- Transformer architecture for time-series regression using self-attention mechanism. 主要パラメータ: num_heads (Number of attention heads), num_layers (Number of transformer layers).ソース: https://github.com/freqtrade/freqtrade/blob/develop/freqai/prediction_models/PyTorchTransformerRegressor.py.
- ReinforcementLearner (Freqtrade)
- Reinforcement learning agent using Stable Baselines3 (PPO/A2C/etc.) for trading decisions. 主要パラメータ: model_type (RL algorithm (PPO, A2C, etc.)), total_timesteps (Training timesteps).ソース: https://github.com/freqtrade/freqtrade/blob/develop/freqai/prediction_models/ReinforcementLearner.py.