数据集获取与加载（Hugging Face datasets）

# 前置依赖: pip install datasets
from datasets import load_dataset

ds = load_dataset("auphong2707/hospital-readmission-risk-data")
print(ds)  # 查看 splits 与字段结构
# 国内网络可先设镜像: import os; os.environ["HF_ENDPOINT"]="https://hf-mirror.com"

git lfs install && git clone https://hf-mirror.com/datasets/auphong2707/hospital-readmission-risk-data

仓库文件结构

• .gitattributes
• README.md
• dataset_info.json
• hospital_readmission_full.csv
• scaler.pkl
• test.csv
• test_demographics.csv
• train.csv
• train_demographics.csv
• validation.csv
• validation_demographics.csv

数据集卡片摘录（源站说明）

Hospital Readmission Risk - Preprocessed Dataset

Dataset Description

This dataset contains preprocessed hospital readmission data for diabetic patients.
The goal is to predict 30-day hospital readmissions to enable proactive interventions.

Dataset Summary

• Total samples: 101,766
• Features: 113
• Target: Binary (0=No readmission, 1=Readmission within 30 days)
• Class distribution: No readmission: 90,409 (88.8%), Readmission: 11,357 (11.2%)

Data Splits

• Training: 71,236 samples
• Validation: 15,265 samples
• Test: 15,265 samples

Preprocessing Applied

1. Missing Value Handling: Median/mode imputation with group-wise strategies
2. Data Validation: Value ranges, data types, and domain constraints checked
3. Outlier Treatment: IQR-based winsorization
4. Feature Engineering:
   • Diagnosis code aggregation into clinical categories
   • Utilization features (group-by statistics)
   • Medication complexity scores
   • Age/BMI categorical buckets
   • Interaction features
5. Encoding: One-hot (low-cardinality) + CV-safe target encoding (high-cardinality)
6. Normalization: StandardScaler applied

Usage

import pandas as pd

# Load full dataset
data = pd.read_csv(''hospital_readmission_full.csv'')
X = data.drop(''target'', axis=1)
y = data[''target'']

# Or load splits
train = pd.read_csv(''splits/train.csv'')
val = pd.read_csv(''splits/validation.csv'')
test = pd.read_csv(''splits/test.csv'')

Citation

Original Dataset: Diabetes 130-US Hospitals for Years 1999-2008
UCI Machine Learning Repository
https://archive.ics.uci.edu/dataset/296/diabetes-130-us-hospitals-for-years-1999-2008

License

See original dataset license from UCI Machine Learning Repository.

数据加载示例（表格/文本类）

import pandas as pd, glob, os

files = (glob.glob(os.path.join(path, "**", "*.csv"), recursive=True)
       + glob.glob(os.path.join(path, "**", "*.tsv"), recursive=True)
       + glob.glob(os.path.join(path, "**", "*.xlsx"), recursive=True))
print("数据文件:", files)
df = pd.read_csv(files[0])
print(df.shape); print(df.columns.tolist()); print(df.head(3))

完整数据卡：huggingface.co/datasets/auphong2707/hospital-readmission-risk-data