synthetic_credit_card_default

# Synthetic Credit Card Default Dataset

### High-fidelity synthetic dataset for financial AI research, created with Syncora.ai

---

## ✅ What's in This Repo?

This repository includes:

- ✅ **Synthetic Credit Card Default Dataset (CSV)** → [Download Here](https://huggingface.co/datasets/syncora/synthetic_credit_card_default/blob/main/UCI_Syncora_Synthetic.csv)
- ✅ **Jupyter Notebook for Analysis & Modeling** → [Open Notebook](https://huggingface.co/datasets/syncora/synthetic_credit_card_default/blob/main/UCI_Syncora_Synthetic.ipynb)
- ✅ **Instructions for generating your own synthetic data using Syncora API**

---

## 📘 About This Dataset

This dataset contains realistic, fully synthetic credit card client records based on the UCI Credit Card Default dataset (2005).
It allows developers and data scientists to perform **credit risk analysis** without using real customer data.
Using **Syncora.ai**, you can also **generate synthetic data** tailored to different modeling scenarios, ensuring **privacy and compliance**.

**Ideal for:**

- Credit risk modeling and analysis
- Binary classification projects
- Explainable AI (XAI) experiments
- Financial ML benchmarking
- Dataset for LLM training (e.g., financial question answering)
- Data science education and prototyping

---

## 🔍 Features

- **Demographics:** Age, sex, education, marital status
- **Credit behavior:** Credit limits, bill amounts, repayment history
- **Target variable:** Default status (`0 = no default, 1 = default`)

---

## 📓 Explore with Our Notebook

A ready-to-run **Jupyter Notebook** demonstrates:

- Loading the dataset from Hugging Face
- Performing credit risk analysis with ML models
- Evaluating performance (accuracy, precision, recall)
- How to **generate synthetic data** for custom scenarios using the Syncora API

👉 [Open the Notebook](https://huggingface.co/datasets/syncora/synthetic_credit_card_default/blob/main/UCI_Syncora_Synthetic.ipynb)

---

## 🚀 Generate Your Own Dataset

Need a dataset for a different scenario?
Create your own **synthetic data for financial AI or dataset for LLM training** with our API:

👉 [Generate synthetic data via Syncora API](https://app.syncora.ai)

---

## ⚡ Quick Start

```python
from datasets import load_dataset

dataset = load_dataset("syncora/synthetic_credit_card_default")
df = dataset["train"].to_pandas()
print(df.head())

# 合成信用卡违约数据集

### 基于Syncora.ai生成的高保真金融AI研究合成数据集

---

## ✅ 本仓库包含内容

本仓库提供以下资源：
- ✅ **合成信用卡违约数据集（CSV格式）** → [点击此处下载](https://huggingface.co/datasets/syncora/synthetic_credit_card_default/blob/main/UCI_Syncora_Synthetic.csv)
- ✅ **用于分析与建模的Jupyter Notebook** → [在线打开Notebook](https://huggingface.co/datasets/syncora/synthetic_credit_card_default/blob/main/UCI_Syncora_Synthetic.ipynb)
- ✅ **使用Syncora API生成自定义合成数据的操作指南**

---

## 📘 数据集简介

本数据集基于2005年UCI信用卡违约数据集生成，包含高度真实的全合成信用卡客户记录，可让开发者与数据科学家无需使用真实客户数据即可开展**信用风险分析**。借助**Syncora.ai**平台，用户还可针对不同建模场景生成定制化合成数据，同时保障数据隐私与合规性。

适用场景包括：
- 信用风险建模与分析
- 二分类任务项目
- 可解释AI（Explainable AI, XAI）实验
- 金融机器学习基准测试
- 大语言模型（Large Language Model, LLM）训练数据集（例如金融问答任务）
- 数据科学教学与原型开发

---

## 🔍 数据集特征

- **人口统计学特征**：年龄、性别、教育程度、婚姻状况
- **信贷行为特征**：信用额度、账单金额、还款历史
- **目标变量**：违约状态（`0 = 未违约，1 = 违约`）

---

## 📓 使用官方Notebook探索数据集

本内置可直接运行的**Jupyter Notebook**可演示以下内容：
- 从Hugging Face加载数据集
- 使用机器学习模型开展信用风险分析
- 评估模型性能（准确率、精确率、召回率）
- 如何通过Syncora API为自定义场景生成合成数据

👉 [在线打开Notebook](https://huggingface.co/datasets/syncora/synthetic_credit_card_default/blob/main/UCI_Syncora_Synthetic.ipynb)

---

## 🚀 生成自定义数据集

需要适配其他场景的数据集？可通过我们的API生成适用于金融AI研究或大语言模型训练的自定义合成数据：

👉 [通过Syncora API生成合成数据](https://app.syncora.ai)

---

## ⚡ 快速入门指南

python
from datasets import load_dataset

dataset = load_dataset("syncora/synthetic_credit_card_default")
df = dataset["train"].to_pandas()
print(df.head())