Anyuhhh/hw2-tabular-automl

AutoML Tabular Model — HW2

Description

Trained with AutoGluon on an augmented tabular split and validated on the original split.

• Best Model: WeightedEnsemble_L2
• Problem Type: binary
• Eval Metric: roc_auc
• Preset: medium_quality

Data

• Train samples: 240
• Test samples: 60
• External (original) samples: 30
• Target column: cover type

Results

• Test (accuracy / f1_macro): 0.7833 / 0.7818
• Original (accuracy / f1_macro): 0.9333 / 0.9321

Top Models (leaderboard head)

model	score_test	score_val	eval_metric	pred_time_test	pred_time_val	fit_time	pred_time_test_marginal	pred_time_val_marginal	fit_time_marginal	stack_level	can_infer	fit_order
WeightedEnsemble_L2	0.899554	0.899554	roc_auc	0.0222933	0.0167627	5.82404	0.00162697	0.000472546	0.0357378	2	True	11
NeuralNetTorch	0.886161	0.886161	roc_auc	0.00805044	0.0069592	3.97173	0.00805044	0.0069592	3.97173	1	True	9
NeuralNetFastAI	0.885045	0.885045	roc_auc	0.0126159	0.00933099	1.81657	0.0126159	0.00933099	1.81657	1	True	7
LightGBMXT	0.863839	0.863839	roc_auc	0.00606823	0.00200415	0.311052	0.00606823	0.00200415	0.311052	1	True	1
RandomForestEntr	0.84654	0.84654	roc_auc	0.0851693	0.0757742	0.609316	0.0851693	0.0757742	0.609316	1	True	4

Hyperparameters (top few)

{
  "note": "Could not fetch: 'TabularPredictor' object has no attribute 'get_hyperparameters'"
}

Usage

import autogluon.tabular as ag
predictor = ag.TabularPredictor.load("path/to/cloned_or_downloaded_repo")
preds = predictor.predict(your_dataframe)

Created: 2025-09-22 01:38