TabCamel

A DataFrame-focused solution for tabular datasets in machine learning workflows.

🎯 Features

• TabularDataset: Comprehensive dataset class with sampling and splitting capabilities
• Data Transformations: Scikit-learn compatible preprocessing transformations
• Multi-source Loading: Support for local files and popular ML repositories
• AutoGluon Integration: Seamless integration with AutoGluon for automated ML

🛠 Installation

pip install tabcamel

🚀 Quick Start

from tabcamel.data.dataset import TabularDataset

# Load a remote dataset
dataset = TabularDataset('iris', task_type='classification')

# Split into train/test sets
train_test = dataset.split('stratified', train_size=0.8)
train_data = train_test['train_set']
test_data = train_test['test_set']

print(train_data)

💽 Dataset Sources

TabCamel supports multiple data sources:

Remote Datasets

• OpenML: 30+ popular datasets ('iris', 'adult', 'titanic', etc.)
• UCI ML Repository: Classic datasets with proper metadata
• scikit-learn: Built-in sklearn datasets ('diabetes', etc.)
• pgmpy: Bayesian network datasets from pgmpy
• bnlearn: datasets from bnlearn

Local Datasets

For local datasets, you have several options:

Option 1: Direct File Path

# Use full path to your dataset
dataset = TabularDataset('/path/to/your/data.csv', task_type='classification')

Option 2: Configure Data Directory

import tabcamel.utils.config as config

# Set up your data directory
local_dataset2path = {
    "local_data": "/path/to/your/data.csv",
}
config.set_local_data_path(local_dataset2path)

# Now use short names
dataset = TabularDataset('local_data', task_type='classification')

💻 Examples

Basic Usage

from tabcamel.data.dataset import TabularDataset

# Remote dataset
dataset = TabularDataset('adult', task_type='classification')

# Local dataset with full path
dataset = TabularDataset('/home/user/data/my_data.csv', task_type='regression')

# Local dataset with configured data directory
dataset = TabularDataset('my_data', task_type='classification')

Data Operations

# Dataset sampling
sample_result = dataset.sample('stratified', sample_size=1000)
sampled_data = sample_result['dataset_sampled']

# Dataset splitting
split_result = dataset.split('stratified', test_size=0.2)
train_set = split_result['train_set']
test_set = split_result['test_set']

# Access properties
print(f"Samples: {dataset.num_samples}")
print(f"Features: {dataset.num_features}")
print(f"Classes: {dataset.num_classes}")

📚 Citation

If you use TabCamel in your research, please cite:

@misc{tabcamel,
  title = {TabCamel: A DataFrame-focused solution for tabular datasets in machine learning workflows},
  author = {Xiangjian Jiang},
  year = {2025},
  howpublished = {\url{https://github.com/SilenceX12138/TabCamel}},
}