ml-autopipeline 0.2.0

Automated ML pipeline with EDA, imbalance detection (SMOTE) and basic model training

ML AutoPipeline

Automated Machine Learning pipeline for beginners.

Features:

• Dataset loading and basic exploratory data analysis (EDA)
• Class imbalance detection and SMOTE sampling
• Model training on Logistic Regression, Random Forest, and SVM
• Performance metrics: Accuracy, Precision, Recall, F1 Score
• Optional extended metrics (confusion matrix, ROC AUC)
• Structured logging (plain or JSON) + timing + optional progress bars
• Config-driven pipeline (YAML/JSON) & CLI flags

Installation

pip install ml-autopipeline
# With optional extras
pip install "ml-autopipeline[all]"     # progress + config
pip install "ml-autopipeline[progress]" # just tqdm
pip install "ml-autopipeline[config]"   # just PyYAML

Quick Start (CLI)

ml-autopipeline --file path/to/data.csv --target target_column --apply_smote

Add verbosity:

ml-autopipeline --file data.csv --target label -v      # INFO
ml-autopipeline --file data.csv --target label -vv     # DEBUG

Extended metrics & progress bar (needs extras installed):

ml-autopipeline --file data.csv --target label --extended-metrics --progress

JSON logs to file:

ml-autopipeline --file data.csv --target label --json-logs --log-file run.log

Use a config file (values can be overridden by CLI):

# config.yml
file: data.csv
target: label
apply_smote: true
extended_metrics: true
progress: true
verbose: 1

Run with config:

ml-autopipeline --config config.yml

Python Usage

from ml_autopipeline import (
    load_data, basic_report, check_imbalance, apply_smote, train_models
)

# df = load_data("your.csv")
report = basic_report(df)
imbalance = check_imbalance(df, "target")
X = pd.get_dummies(df.drop(columns=["target"]))
y = df["target"]
if imbalance["is_imbalanced"]:
    X, y = apply_smote(X, y)
results = train_models(X, y, extended_metrics=True, show_progress=True)
print(results)

Provided Functions

Function	Description
load_data(path)	Load CSV into a pandas DataFrame
basic_report(df)	Return shape, columns, missing values, dtypes, head
check_imbalance(df, target)	Report class distribution and imbalance flag
apply_smote(X, y)	Oversample minority classes using SMOTE / fallback strategy
train_models(X, y, ...)	Train models and return metrics (+ optional extended metrics)

Extended metrics keys (when enabled):

• confusion_matrix: 2D list
• roc_auc or roc_auc_ovr_weighted (if probabilities available)

Logging & Timing

• Default log level: WARNING
• -v -> INFO, -vv -> DEBUG
• --json-logs produces JSON per line
• --log-file FILE duplicates logs to file
• Timing for major steps included (split, fit per model)

Config Precedence

1. CLI arguments (if provided)
2. Config file values

Development Setup

python -m venv .venv
.venv\Scripts\activate
pip install -e .[dev,all]
pytest

Build & Publish

pip install build twine
python -m build
python -m twine upload --repository testpypi dist/*
# After verification
python -m twine upload dist/*

License

MIT