pyflowml 1.0.0

AutoML library for the full ML lifecycle — train, evaluate, visualise, and save

PyFlowML 🚀

A scalable, intelligent Python ML library for the full machine learning lifecycle.

PyFlowML handles everything from raw, potentially large data to a trained, evaluated, and saved model — automatically.

---

Features

Feature	Description
🧠 AutoML	Trains multiple models in parallel, picks the best automatically
📊 Data Profiler	Detects problem type, missing data, skewness, and outliers
⚙️ Smart Pipeline	Cached preprocessing: no redundant recomputation
📦 Memory Optimizer	Downcasts dtypes — saves 40–60% RAM
🚀 Parallel Training	All models train simultaneously via joblib
⏱️ Time Budget	AutoClassifier(time_limit=60) stops when budget exceeded
📈 Visualization	Dark-themed ROC, confusion matrix, learning curves
💾 Versioned Saves	Models saved with metadata (date, metrics, features)
📝 NLP Utilities	TF-IDF, Bag-of-Words, stopword removal, lemmatization
📡 Monitoring	Per-step time and memory tracking

---

Installation

pip install -e .

Or install from requirements:

pip install -r requirements.txt

---

Quickstart

One-line AutoML Engine

import pandas as pd
from pyflowml.core.engine import PyFlowEngine

df = pd.read_csv("titanic.csv")
engine = PyFlowEngine(df, target="Survived", time_limit=120)
engine.run()

Manual Step-by-Step

from pyflowml.data import DataLoader, DataCleaner, DataSplitter, MemoryOptimizer
from pyflowml.preprocessing import SmartPipeline
from pyflowml.models import AutoClassifier
from pyflowml.evaluation import Reporter
from pyflowml.visualization import ModelViz
from pyflowml.utils import ModelSaver

# Load & optimize
df = DataLoader.from_csv("data.csv")
df = MemoryOptimizer.reduce(df)

# Clean
df = DataCleaner(df).handle_nulls().remove_outliers().remove_duplicates().result()

# Split
X_train, X_test, y_train, y_test = DataSplitter(df, target="label").split()

# Preprocess
pipe = SmartPipeline()
X_train = pipe.fit_transform(X_train, y_train)
X_test  = pipe.transform(X_test)

# Train best model
clf = AutoClassifier(metric="f1", time_limit=60)
clf.fit(X_train, y_train)
clf.leaderboard()

# Evaluate
Reporter.classification(clf, X_test, y_test)

# Visualize
ModelViz.confusion_matrix(clf, X_test, y_test)
ModelViz.roc_curve(clf, X_test, y_test)

# Save
ModelSaver.save(clf, "my_model", metadata={"f1": clf.best_score_})

---

Modules

pyflowml/
├── core/          # Engine, Profiler, Optimizer (brain of the system)
├── data/          # DataLoader, DataCleaner, DataSplitter, MemoryOptimizer
├── preprocessing/ # Scaler, FeatureSelector, SmartPipeline
├── models/        # AutoClassifier, AutoRegressor, AutoClusterer
├── evaluation/    # Reporter, CrossValidator
├── tuning/        # HyperTuner
├── visualization/ # Plotter, ModelViz
├── monitoring/    # StepTracker, Logger
├── utils/         # ModelSaver
└── nlp/           # TextCleaner, Vectorizer

---

Running Tests

pytest tests/ -v

---

License

MIT