outlier-toolkit 0.1.3

A Python library for identifying and handling outliers

🧰 Outlier Toolkit 🛠️

A standalone Python library for detecting, handling, and transforming outliers in numeric and categorical data.
No external dependencies required.

---

📜 License

This project is licensed under the Apache License 2.0. See the LICENSE file for more details.

---

📊 Features

1. Outlier Detection

• Z-score Detection: Identify extreme values based on standard deviation.
• IQR Detection: Detect outliers using the interquartile range (Q1, Q3).

2. Outlier Handling Techniques

• Remove Outliers: Drop outlier values from datasets.
• Replace Outliers: Replace outliers with mean, median, or most frequent values.

3. Winsorization

• Standard Winsorization: Cap extreme values at a fixed percentile.
• Adaptive Quartiles: Replace low/high outliers using Q1 and Q3.
• Adaptive Inliers: Replace low/high outliers using nearest inlier values (custom method).

4. Binning

• Equal Width Binning: Divide numeric range into equal-width intervals.
• Equal Frequency Binning: Divide data so each bin has approximately the same number of values.
• Auto Binning (Outlier-based): Automatically separate low/high outliers and inliers using IQR.

---

🔧 Installation

pip install outlier-toolkit

No external libraries required. Compatible with Python 3.7+.

---

🧮 Usage

from outlier.i_outlier.Zscore import detect_outliers_zscore
from outlier.i_outlier.IQR import detect_outliers_iqr
from outlier.outlierTech.remove import remove_outliers
from outlier.outlierTech.replace import replace_outliers
from outlier.outlierTech.winsorization.standard import winsorize_standard
from outlier.outlierTech.winsorization.adaptive import winsorize_quartiles
from outlier.outlierTech.winsorization.adaptive import winsorize_inliers
from outlier.outlierGroup.binning import eq_width_bin
from outlier.outlierGroup.binning import eq_freq_bin
from outlier.outlierGroup.binning import custom_binning



# Sample Test data
numeric_data = [1, 2, 85, 95, 65, 75, 53, 67, 87, 89, 93, 1001, 1027, 3018]
categorical_data = ["Male", "Female", "Male", "Male", "Unknown", "Unknown", "Other"]

#Detection
print("=== Zscore Detection ===")
print(detect_outliers_zscore(numeric_data))

print("\n=== IQR Detection ===")
print(detect_outliers_iqr(numeric_data))

#Handling
print("\n=== Remove Outliers ===")
print(remove_outliers(numeric_data, method="IQR"))

print("\n=== Replace Outliers (auto-detect) ===")
print(replace_outliers(numeric_data, method="IQR"))
print(replace_outliers(categorical_data, method="IQR"))

#Winsorization
print("\n=== Winsorization (Standard 5%) ===")
print(winsorize_standard(numeric_data[:]))

print("\n=== Winsorization (Adaptive Quartiles) ===")
print(winsorize_quartiles(numeric_data[:]))

print("\n=== Winsorization (Adaptive Inliers) ===")
print(winsorize_inliers(numeric_data[:]))

#Binning
print("\n=== Binning (Equal Width Binning) ===")
print(eq_width_bin(numeric_data[:]))

print("\n=== Binning (Equal Width Binning) ===")
print(eq_freq_bin(numeric_data[:]))

print("\n=== Binning (Equal Width Binning) ===")
print(custom_binning(numeric_data[:]))

---

📝 Notes

• Works for numeric and categorical data.
• All functions are standalone and do not require external libraries.
• Custom winsorization allows mapping outliers to nearest inliers for more controlled transformations.

---

👩‍💻 Author

Irene Betsy D

---