From eb75ddf1e4d32f01053d78d42e65e188cb1cddef Mon Sep 17 00:00:00 2001
From: stingray <114566779+stingraypark@users.noreply.github.com>
Date: Thu, 8 Jan 2026 22:33:30 +0900
Subject: [PATCH] Add introduction and purpose to README.md

Added an introduction and purpose section for GPC Encoder, outlining limitations of common encoding methods and the advantages of GPC Encoder.
---
 README.md | 31 +++++++++++++++++++++++++++++++
 1 file changed, 31 insertions(+)
 create mode 100644 README.md

diff --git a/README.md b/README.md
new file mode 100644
index 0000000..20570a8
--- /dev/null
+++ b/README.md
@@ -0,0 +1,31 @@
+# GPC Encoder (Geometry Preserving Category Encoder)
+
+## Introduction & Purpose
+
+In ML/DL preprocessing, categorical data is commonly encoded using **one-hot**, **label**, or **target mean encoding**, each of which has clear limitations:
+
+- **One-Hot Encoding**
+  - Causes a rapid increase in dimensionality for high-cardinality features
+  - Increases model complexity and memory usage
+
+- **Label Encoding**
+  - Introduces artificial ordinal relationships between categories
+  - Can mislead linear or distance-based models
+
+- **Target Mean Encoding**
+  - Prone to **data leakage**, as global target statistics influence encoding
+  - Unstable for rare categories or small datasets, leading to bias
+
+The purpose of **GPC Encoder** is to address all of these issues by designing an encoding method that:
+
+1. Does **not impose any ordering** on categories  
+2. Does **not drastically increase dimensionality**  
+3. **Avoids data leakage** by relying only on training data statistics  
+
+---
+
+## Requirements
+
+- **Python 3.11.4**
+- See `requirements.txt` for full dependency details
+