Step #2

Generating Out-of-Fold (OOF) Target Values using AutoGluon

Last update: August 14, 2025

AI Assistance: Claude.AI (Anthropic) is used for documentation, code restructuring, and performance optimization.

Copyright (C) 2025 Sukanta Basu

This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with this program. If not, see https://www.gnu.org/licenses/.

Overall Strategy

Step 1: Preprocess and engineer new features.

Step 2: Use AutoGluon to generate OOF predictions for each target separately. These predictions will be used as additional input features in steps 3 and 4.

Step 3: Train the RealMLP model with processed input (step 1) + ten AutoGluon-OOFs (step 2). These additional features will capture the correlation among targets effectively.

Step 4: Similar to step 3 except use the TabPFN (v2) model.

Step 5: Combine predictions from RealMLP (step 3) and TabPFN (step 4).

Imports

import numpy as np
import pandas as pd
import os
import random
import warnings

from autogluon.tabular import TabularPredictor

Set Random Seeds

random.seed(7)
np.random.seed(7)

User Input

# AutoGluon quality preset
quality_preset = 'best_quality'

# AutoGluon training time (in seconds)
maxTime = 21600

# Number of input features
nFeatures = 65

# Number of target variables
nTargets = 10

Input & Output Directories

ROOT_DIR = '/data/Sukanta/Works_AIML/2025_SHELL_FuelProperty/'
DATA_DIR = ROOT_DIR + 'DATA/'
ExtractedDATA_DIR = ROOT_DIR + 'ExtractedDATA/'
Tuning_DIR = ROOT_DIR + 'Models/AutoGluon-OOF/'

# Create directory if it doesn't exist
os.makedirs(Tuning_DIR, exist_ok=True)

Load Processed Training and Testing Data

df_XyTrnVal_org = pd.read_csv(ExtractedDATA_DIR + 'train_processed.csv')
nSamples_TrnVal = df_XyTrnVal_org.shape[0]

df_XTst = pd.read_csv(ExtractedDATA_DIR + 'test_processed.csv')
nSamples_Tst = df_XTst.shape[0]

print(f"Training data shape: {df_XyTrnVal_org.shape}")
print(f"Test data shape: {df_XTst.shape}")

# Extract input features
XTrnVal = df_XyTrnVal_org.iloc[:, 0:nFeatures]

Iterative Single-target Training using AutoGluon

# Initialize predictions array
yTrnVal_OOF = np.zeros((nSamples_TrnVal, nTargets))
yTst = np.zeros((nSamples_Tst, nTargets))

for target in range(nTargets):
    print(f"\n--- Target {target + 1}/{nTargets} ---")

    # Extract single target from possible nTargets
    yTrnVal = df_XyTrnVal_org.iloc[:, nFeatures + target]

    # Create training dataframe with features and target
    train_data = XTrnVal.copy()
    train_data[f'target_{target}'] = yTrnVal

    # Create unique file path for each target
    target_path = os.path.join(Tuning_DIR, f'target_{target + 1}')
    os.makedirs(target_path, exist_ok=True)

    # Initialize TabularPredictor from AutoGluon
    predictor = TabularPredictor(
        label=f'target_{target}',
        path=target_path,
        eval_metric='mean_absolute_percentage_error',
        problem_type='regression'
    )

    # Train the model
    print("Starting AutoGluon training...")
    predictor.fit(
        train_data,
        time_limit=maxTime,
        presets=quality_preset,
        verbosity=2,
        auto_stack=False,
        dynamic_stacking=False,
        num_bag_folds=8,
        num_bag_sets=5,
        num_stack_levels=2,
        use_bag_holdout=False,
        fit_strategy="sequential",
        ag_args_ensemble={'fold_fitting_strategy': "parallel_local"},
        ds_args={'enable_ray_logging': False}
    )

    print("\n Model Leaderboard:")
    leaderboard = predictor.leaderboard(silent=True)
    print(leaderboard.sort_values("score_val", ascending=False).head())

    # OOF predictions based on training set
    yTrnVal_OOF[:, target] = predictor.predict_oof()

    # Make predictions on test set
    yTst[:, target] = predictor.predict(df_XTst)
    print(f"Test predictions generated for target {target + 1}")

    # Clean up predictor to free memory
    del predictor

Save Results

print("\n=== SAVING RESULTS ===")

# Create dataframes
df_AG_yTrnVal_OOF = pd.DataFrame()
df_AG_yTst = pd.DataFrame()

# Add prediction columns
for i in range(nTargets):
    df_AG_yTrnVal_OOF[f'AG-BlendProperty{i + 1}'] = yTrnVal_OOF[:, i]
    df_AG_yTst[f'AG-BlendProperty{i+1}'] = yTst[:, i]

# Save predictions
AG_OOF_file = os.path.join(ExtractedDATA_DIR, f'AutoGluon_{maxTime}_OOF.csv')
df_AG_yTrnVal_OOF.to_csv(AG_OOF_file, index=False)

AG_Tst_file = os.path.join(ExtractedDATA_DIR, f'AutoGluon_{maxTime}_Tst.csv')
df_AG_yTst.to_csv(AG_Tst_file, index=False)

print(f"AutoGluon training completed!")