alpha_lab/cta_1d/03_baseline_xgb.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# CTA 1D Baseline XGBoost Model\n",
    "\n",
    "Train and evaluate a baseline XGBoost model for CTA 1-day return prediction.\n",
    "\n",
    "**Purpose**: Establish a baseline performance benchmark with standard configuration."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import json\n",
    "from datetime import datetime\n",
    "\n",
    "from qshare.data.pandas.cta_1d import load_dataset\n",
    "from qshare.algo.learning.cta_trainer import CTAXGBTrainer\n",
    "from qshare.eval.cta.backtest import CTABacktester\n",
    "\n",
    "import sys\n",
    "sys.path.insert(0, '../')\n",
    "from common.plotting import setup_plot_style, plot_ic_series, plot_cumulative_returns\n",
    "from common.paths import create_experiment_dir\n",
    "from src.labels import get_blend_weights, describe_blend_config\n",
    "\n",
    "setup_plot_style()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1. Configuration\n",
    "\n",
    "Edit this cell to modify experiment parameters."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "CONFIG = {\n",
    "    # Experiment\n",
    "    'experiment_name': 'baseline_xgb',  # Will be appended with timestamp\n",
    "    \n",
    "    # Date ranges\n",
    "    'dt_range': ['2020-01-01', '2024-12-31'],\n",
    "    'train_range': ['2020-01-01', '2022-12-31'],\n",
    "    'test_range': ['2023-01-01', '2024-12-31'],\n",
    "    'fit_range': ['2020-01-01', '2021-06-30'],  # For normalization fitting\n",
    "    \n",
    "    # Data\n",
    "    'feature_sets': ['alpha158', 'hffactor'],\n",
    "    'return_type': 'o2c_twap1min',\n",
    "    'normalization': 'dual',\n",
    "    'blend_weights': None,  # Use default [0.2, 0.1, 0.3, 0.4] or specify name/list\n",
    "    'weight_factors': {'positive': 1.0, 'negative': 2.0},\n",
    "    \n",
    "    # Model\n",
    "    'xgb_params': {\n",
    "        'booster': 'gblinear',\n",
    "        'eta': 0.5,\n",
    "        'lambda_reg': 0.1,\n",
    "        'num_round': 20,\n",
    "    },\n",
    "    \n",
    "    # Backtest\n",
    "    'backtest_params': {\n",
    "        'num_trades': 4,\n",
    "        'signal_dist': 'normal',\n",
    "        'pos_weight': True,\n",
    "    },\n",
    "    \n",
    "    # Output\n",
    "    'save_results': True,\n",
    "}\n",
    "\n",
    "print(\"Configuration:\")\n",
    "print(f\"  Experiment: {CONFIG['experiment_name']}\")\n",
    "print(f\"  Train: {CONFIG['train_range'][0]} to {CONFIG['train_range'][1]}\")\n",
    "print(f\"  Test: {CONFIG['test_range'][0]} to {CONFIG['test_range'][1]}\")\n",
    "print(f\"  Blend: {describe_blend_config(CONFIG['blend_weights'] or 'default')}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. Load Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(\"Loading dataset...\")\n",
    "df_full = load_dataset(\n",
    "    dt_range=CONFIG['dt_range'],\n",
    "    return_type=CONFIG['return_type'],\n",
    "    normalization=CONFIG['normalization'],\n",
    "    feature_sets=CONFIG['feature_sets'],\n",
    "    fit_range=CONFIG['fit_range'],\n",
    "    weight_factors=CONFIG['weight_factors'],\n",
    "    blend_weights=CONFIG['blend_weights'],\n",
    ")\n",
    "\n",
    "print(f\"\\nDataset shape: {df_full.shape}\")\n",
    "print(f\"Columns: {len(df_full.columns)}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Split train/test\n",
    "df_train = df_full.loc[CONFIG['train_range'][0]:CONFIG['train_range'][1]]\n",
    "df_test = df_full.loc[CONFIG['test_range'][0]:CONFIG['test_range'][1]]\n",
    "\n",
    "print(f\"Train: {df_train.shape}\")\n",
    "print(f\"Test:  {df_test.shape}\")\n",
    "\n",
    "# Get feature columns\n",
    "feature_cols = [c for c in df_train.columns\n",
    "                if c.startswith(('alpha158_', 'hf_', 'f_'))]\n",
    "print(f\"\\nFeatures: {len(feature_cols)}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3. Train Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(\"Training XGBoost model...\")\n",
    "print(f\"  Params: {CONFIG['xgb_params']}\")\n",
    "\n",
    "trainer = CTAXGBTrainer(**CONFIG['xgb_params'])\n",
    "\n",
    "trainer.fit(\n",
    "    df_train,\n",
    "    feature_cols=feature_cols,\n",
    "    target_col='label',\n",
    "    weight_col='weight'\n",
    ")\n",
    "\n",
    "print(\"\\nTraining complete!\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Feature importance\n",
    "importance = trainer.get_feature_importance()\n",
    "print(\"\\nTop 10 Features:\")\n",
    "print(importance.head(10))\n",
    "\n",
    "# Plot\n",
    "fig, ax = plt.subplots(figsize=(10, 6))\n",
    "importance.head(20).plot(kind='barh', ax=ax)\n",
    "ax.set_title('Top 20 Feature Importance')\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 4. Generate Predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(\"Generating predictions on test set...\")\n",
    "df_signal = trainer.predict(df_test)\n",
    "\n",
    "print(f\"\\nSignal statistics:\")\n",
    "print(df_signal.describe())\n",
    "\n",
    "# Plot signal distribution\n",
    "fig, axes = plt.subplots(1, 2, figsize=(14, 4))\n",
    "\n",
    "df_signal.hist(bins=100, ax=axes[0], edgecolor='black')\n",
    "axes[0].set_title('Signal Distribution')\n",
    "axes[0].axvline(x=0, color='red', linestyle='--')\n",
    "\n",
    "signal_by_date = df_signal.groupby(level=0).mean()\n",
    "axes[1].plot(signal_by_date.index, signal_by_date.values)\n",
    "axes[1].set_title('Mean Signal by Date')\n",
    "axes[1].axhline(y=0, color='red', linestyle='--')\n",
    "\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 5. Evaluate with Backtest"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(\"Running backtest...\")\n",
    "\n",
    "returns = df_test['return'] if 'return' in df_test.columns else df_test['label']\n",
    "\n",
    "backtester = CTABacktester(**CONFIG['backtest_params'])\n",
    "results = backtester.run(returns, df_signal)\n",
    "\n",
    "summary = backtester.summary()\n",
    "print(\"\\nBacktest Summary:\")\n",
    "for key, value in summary.items():\n",
    "    if isinstance(value, float):\n",
    "        print(f\"  {key}: {value:.4f}\")\n",
    "    else:\n",
    "        print(f\"  {key}: {value}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# IC Analysis\n",
    "ic_by_date = results.groupby(results.index.get_level_values(0))['ic'].mean()\n",
    "\n",
    "fig = plot_ic_series(ic_by_date, title=\"IC Over Time (Test Set)\")\n",
    "plt.show()\n",
    "\n",
    "print(f\"\\nIC Statistics:\")\n",
    "print(f\"  Mean: {ic_by_date.mean():.4f}\")\n",
    "print(f\"  Std:  {ic_by_date.std():.4f}\")\n",
    "print(f\"  IR:   {ic_by_date.mean() / ic_by_date.std():.4f}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Cumulative returns\n",
    "daily_returns = results.groupby(results.index.get_level_values(0))['pos_return'].mean()\n",
    "\n",
    "fig = plot_cumulative_returns(daily_returns, title=\"Cumulative Strategy Returns\")\n",
    "plt.show()\n",
    "\n",
    "total_return = (1 + daily_returns).prod() - 1\n",
    "annual_return = (1 + total_return) ** (252 / len(daily_returns)) - 1\n",
    "sharpe = daily_returns.mean() / daily_returns.std() * np.sqrt(252)\n",
    "\n",
    "print(f\"\\nReturn Statistics:\")\n",
    "print(f\"  Total Return: {total_return:.2%}\")\n",
    "print(f\"  Annual Return: {annual_return:.2%}\")\n",
    "print(f\"  Sharpe Ratio: {sharpe:.2f}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 6. Save Results\n",
    "\n",
    "Save model, predictions, and metrics for later analysis."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "if CONFIG['save_results']:\n",
    "    # Create output directory\n",
    "    output_dir = create_experiment_dir('cta_1d', CONFIG['experiment_name'])\n",
    "    print(f\"Saving results to: {output_dir}\")\n",
    "    \n",
    "    # Save config\n",
    "    with open(output_dir / 'config.json', 'w') as f:\n",
    "        json.dump(CONFIG, f, indent=2, default=str)\n",
    "    \n",
    "    # Save model\n",
    "    trainer.save_model(str(output_dir / 'model.pkl'))\n",
    "    \n",
    "    # Save feature importance\n",
    "    importance.to_csv(output_dir / 'feature_importance.csv')\n",
    "    \n",
    "    # Save predictions\n",
    "    df_signal.to_csv(output_dir / 'predictions.csv')\n",
    "    \n",
    "    # Save backtest results\n",
    "    results.to_csv(output_dir / 'backtest_results.csv')\n",
    "    \n",
    "    # Save summary\n",
    "    with open(output_dir / 'summary.json', 'w') as f:\n",
    "        json.dump(summary, f, indent=2, default=str)\n",
    "    \n",
    "    print(\"\\nFiles saved:\")\n",
    "    for f in output_dir.iterdir():\n",
    "        print(f\"  - {f.name}\")\n",
    "else:\n",
    "    print(\"Results not saved (save_results=False)\")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "name": "python",
   "version": "3.8.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}