preplify 1.0.3

Modular, professional Python library for tabular data preprocessing with auto ML-ready pipelines.

🚀 Preplify

The Last Preprocessing Library You'll Ever Need

Preprocess any tabular dataset in one line — clean, encode, scale, engineer, and ML-ready.

# From messy raw data to ML-ready in ONE line
df_clean = auto_prep(df)

---

📌 What is Preplify?

Preplify is a modular, professional Python library for tabular data preprocessing. Whether you're a beginner who wants everything handled automatically, or an advanced user who wants full control over every step — Preplify has you covered.

No more writing 100 lines of boilerplate preprocessing code. Preplify does it all.

---

✨ Features at a Glance

Module	What it does
🧹 Cleaning	Remove duplicates, empty rows, fix column names
❓ Missing Values	Mean / Median / Mode / Drop / Constant strategies
📊 Outlier Removal	IQR and Z-score based detection
🔠 Encoding	One-Hot and Label encoding
📏 Scaling	Standard, Min-Max, Robust scaling
🔄 Transformation	Log and Power transforms
⚙️ Feature Engineering	Interaction, Ratio, and DateTime features
🎯 Feature Selection	Correlation filter, Variance filter, Importance selector
📉 Dimensionality Reduction	PCA
🔍 Profiling	Full data report + summary stats
💡 Recommender	Auto-suggests best preprocessing steps
🤖 AutoML	Preprocess + train baseline model in one call
🔧 Pipeline	Custom modular pipeline with full control

---

⚡ Installation

pip install preplify

Requirements:

pip install pandas numpy scikit-learn

Python 3.8+ required

---

🚀 Quick Start

One-Line Preprocessing

import pandas as pd
from preplify import auto_prep

df = pd.read_csv("your_dataset.csv")
df_clean = auto_prep(df)
print(df_clean.shape)

See What Your Data Needs First

from preplify import recommend_preprocessing, data_report

_ = recommend_preprocessing(df)   # get smart suggestions
_ = data_report(df)                # full data overview

Train a Baseline Model Instantly

from preplify import automl_prep

X, model, score = automl_prep(df, target="Survived", task="classification")
print(f"Accuracy: {score:.4f}")

---

🔧 Full API Reference

📦 Load Demo Data

from preplify.datasets.demo_datasets import (
    load_sample_classification,   # fake classification dataset
    load_sample_regression,       # fake regression dataset
    load_titanic                  # real Titanic CSV (needs internet)
)

df = load_sample_classification()
df = load_sample_regression()
df = load_titanic()

---

🔍 Profiling

from preplify import data_report, recommend_preprocessing

_ = data_report(df)             # shape, missing, dtypes, duplicates
_ = recommend_preprocessing(df) # smart preprocessing suggestions

---

🧹 Cleaning

from preplify import remove_duplicates, remove_empty_rows, standardize_columns

df = standardize_columns(df)   # "First Name" → "first_name"
df = remove_duplicates(df)     # drop repeated rows
df = remove_empty_rows(df)     # drop fully empty rows

---

❓ Missing Values

from preplify import handle_missing

df = handle_missing(df, strategy="mean")                    # fill with average
df = handle_missing(df, strategy="median")                  # fill with middle value
df = handle_missing(df, strategy="mode")                    # fill with most frequent
df = handle_missing(df, strategy="drop")                    # drop rows with missing
df = handle_missing(df, strategy="constant", fill_value=0)  # fill with 0

---

📊 Outlier Removal

from preplify import remove_outliers, remove_outliers_iqr, remove_outliers_zscore

df = remove_outliers(df, method="iqr")          # IQR method (recommended)
df = remove_outliers(df, method="zscore")       # Z-score method
df = remove_outliers_iqr(df)                    # direct IQR
df = remove_outliers_zscore(df, threshold=3)    # direct Z-score

---

🔠 Encoding

from preplify import encode_features, onehot_encode, label_encode

df = encode_features(df, method="onehot")   # "Male/Female" → 2 columns
df = encode_features(df, method="label")    # "Male/Female" → 0/1
df = onehot_encode(df)                      # direct one-hot
df = label_encode(df)                       # direct label encoding

---

📏 Scaling

from preplify import scale_features, minmax_scale, robust_scale

df = scale_features(df, method="standard")  # mean=0, std=1
df = scale_features(df, method="minmax")    # values between 0 and 1
df = scale_features(df, method="robust")    # outlier-resistant scaling
df = minmax_scale(df)                       # shortcut minmax
df = robust_scale(df)                       # shortcut robust

---

🔄 Transformation

from preplify import log_transform, power_transform

df = log_transform(df)                           # log1p — fix skewed data
df = power_transform(df, method="yeo-johnson")   # works with negatives too
df = power_transform(df, method="box-cox")       # positive values only

---

⚙️ Feature Engineering

from preplify import (
    auto_feature_engineering,
    interaction_features,
    ratio_features,
    extract_date_features
)

df = auto_feature_engineering(df)                      # auto interaction + ratio
df = interaction_features(df)                          # age * income → age_x_income
df = ratio_features(df)                                # income / age → income_ratio_age
df = extract_date_features(df, date_columns=["date"])  # year, month, day, weekday, is_weekend

---

🎯 Feature Selection

from preplify import correlation_filter, variance_filter, importance_selector

df = correlation_filter(df, threshold=0.9)  # drop columns that are 90%+ similar
df = variance_filter(df, threshold=0.01)    # drop near-constant columns
X  = importance_selector(X, y, top_n=10)   # keep top 10 by Random Forest importance

---

📉 Dimensionality Reduction

from preplify import apply_pca

df = apply_pca(df, n_components=0.95)   # keep 95% of variance
df = apply_pca(df, n_components=3)      # reduce to exactly 3 components

---

✂️ Train-Test Split

from preplify import split_dataset

X_train, X_test, y_train, y_test = split_dataset(X, y, test_size=0.2)

---

🔧 Custom Pipeline

Full control over every single step:

from preplify import PreplifyPipeline

pipe = PreplifyPipeline(
    missing_strategy="median",     # mean / median / mode / drop / constant
    encoding="onehot",             # onehot / label
    scaling="robust",              # standard / minmax / robust
    outlier_method="iqr",          # iqr / zscore / None
    feature_engineering=True       # True / False
)

df_clean = pipe.fit_transform(df)

---

🤖 AutoML — Preprocess + Train in One Call

from preplify import automl_prep

# Classification
X, model, score = automl_prep(df, target="Survived", task="classification")
print(f"Accuracy: {score:.4f}")

# Regression
X, model, score = automl_prep(df, target="Price", task="regression")
print(f"R² Score: {score:.4f}")

Task	Model Used	Score Metric
classification	Logistic Regression	Accuracy
regression	Ridge Regression	R² Score

---

🌍 Real Dataset Example (Titanic)

import pandas as pd
from preplify import auto_prep, recommend_preprocessing
from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score

# Load
df = pd.read_csv("titanic.csv")

# Explore
_ = recommend_preprocessing(df)

# Preprocess
target = df["Survived"]
df_clean = auto_prep(df.drop(columns=["Survived"]))

# Train Random Forest
X_train, X_test, y_train, y_test = train_test_split(df_clean, target, test_size=0.2, random_state=42)
model = RandomForestClassifier()
model.fit(X_train, y_train)
print(f"Accuracy: {accuracy_score(y_test, model.predict(X_test)):.4f}")

---

🗂️ Project Structure

preplify/
│
├── preplify/
│   ├── core/              # Pipeline, auto_prep, automl_prep
│   ├── cleaning/          # Duplicates, empty rows, column names
│   ├── missing/           # Missing value strategies
│   ├── outliers/          # IQR and Z-score detection
│   ├── encoding/          # One-hot and label encoding
│   ├── scaling/           # Standard, MinMax, Robust
│   ├── transformation/    # Log and Power transforms
│   ├── feature_engineering/  # Interaction, ratio, date features
│   ├── feature_selection/ # Correlation, variance, importance
│   ├── reduction/         # PCA
│   ├── split/             # Train-test split
│   ├── profiling/         # Data report, summary stats
│   ├── recommender/       # Preprocessing recommendations
│   ├── datasets/          # Demo datasets
│   └── utils/             # Validators, logging, helpers
│
├── examples/              # Ready-to-run example scripts
├── tests/                 # Unit tests
├── docs/                  # Documentation
├── setup.py
└── README.md

---

🔕 Disable Logs

By default Preplify logs every step. To turn off:

import logging
logging.getLogger("preplify").setLevel(logging.WARNING)

---

🧪 Run Tests

pip install pytest
pytest tests/

---

📄 License

MIT License — free to use, modify, and distribute.

---

👨‍💻 Author

Muhammad Hussnain

• 📧 [e-mail]
• ⭐ GitHub Repo
• 🐙 GitHub Profile
• 💼 LinkedIn Profile

---

If Preplify saved you time, give it a ⭐ on GitHub!