AutoFlowML

Automated data preprocessing and model selection for production and analytics.

Overview

AutoFlowML automates the tedious parts of machine learning:

Data Cleaning (missing values, outliers, duplicates)

Feature Engineering (encoding, scaling, selection)

Model Selection (AutoML with optimized hyperparameters)

Built for data scientists who want to focus on insights, not boilerplate code.

---

Installation

pip install autoflowml

---

Getting Started

from autoflowml import CleanIt, NullFixer, run_tiny_automl
import pandas as pd

# Load your dataset
df = pd.read_csv("data.csv")

# Step 1: Clean the data
df_clean = CleanIt(df).full_clean()

# Step 2: Fix missing values
df_imputed = NullFixer(df_clean).nullfix_knn(n_neighbors=5)

# Step 3: Train the best model
run_tiny_automl(df_imputed, target_column="target", problem_type="regression")

---

Show Off More Features!

AutoFlowML is built for production. Here's how to remove outliers, encode features, and save models.

from autoflowml import AutoOutlier, CategoricalMaster
from joblib import dump, load

# Step 1: Outlier removal
df_no_outliers = AutoOutlier(method="isolation_forest").fit_transform(df_imputed)

# Step 2: Categorical encoding
encoded_df = CategoricalMaster(df_no_outliers, target_column="target").encode_auto()

# Step 3: Train the model
model = run_tiny_automl(encoded_df, target_column="target", problem_type="classification")

# Step 4: Save the model
dump(model, "best_model.pkl")

# Step 5: Load and predict
model = load("best_model.pkl")
predictions = model.predict(encoded_df.drop("target", axis=1))

---

3rd Party Integrations

AutoFlowML plays well with popular ML libraries:

• evalml for automated model selection and hyperparameter tuning
• category_encoders for powerful encoding techniques
• Fully compatible with scikit-learn pipelines

---

Feature Highlights

Analytics + Cleaning

• Rename messy columns
• Remove duplicates
• Fix null values with KNN, MICE, or time-aware methods

Outlier Detection

• Supports Z-score, IQR, Isolation Forest

Encoding Support

• One-Hot, Binary, and Target encoding built-in

AutoML

• Regression and Classification
• Feature selection and model tuning with evalml

Ready for Deployment

• Serialize models using joblib
• Predict on individual rows or full DataFrames

---

Feature Learning (Advanced)

You can use deep learning to extract useful features, then combine them with traditional models for superior performance. Use feature_learning=True in run_tiny_automl() (coming soon).

---

Categorical Ensembling (Coming Soon)

Train one model per category with train_categorical_ensemble(), perfect for use-cases like one model per region/store/customer.

---

Documentation

Full documentation coming soon. Until then, refer to examples in this README or explore the API directly.

---

What This Project Automates

• Column cleanup, outlier detection, and data imputation
• Encoding categorical variables
• Automated model selection and evaluation
• Feature importance and selection
• Serialization for production

---

Running Tests

If contributing or debugging, run:

pytest tests/

---

Why AutoFlowML?

• Saves hours of preprocessing
• Production-ready pipelines
• Transparent, extensible design
• Lightning-fast single-row predictions