kastor-xgb-demo-nq-ta-v1

This is an XGBoost model trained to classify 15-minute forward price movement for NQ futures. It was optimized using Optuna and is part of the Kastor AI project.

Model Details

• Algorithm: XGBoost (XGBClassifier)
• Data: 1-minute NQ futures OHLCV data from the mdelcristo/NQ-F_1min_OHLCV_Parquet dataset.
• Features: ['RSI', 'ATR', 'MACD', 'MACD_signal']
• Target: 15-minute forward price movement classified as Up (1), Flat (0), or Down (-1).

Training

The model was tuned with Optuna over 20 trials and trained on data up to 2023-01-01.

Evaluation Results (Test Set: 2023-Present)

          precision    recall  f1-score   support

Down (-1) 0.4003 0.2266 0.2894 359156 Flat (0) 0.1788 0.4929 0.2624 157355 Up (1) 0.4350 0.2996 0.3549 387498

accuracy                         0.3043    904009

macro avg 0.3380 0.3397 0.3022 904009 weighted avg 0.3766 0.3043 0.3128 904009