training-data

What problem does it solve?

This Skill addresses the critical challenge of preparing high-quality training data for machine learning models, ensuring better model performance and reliability.

Core Features & Use Cases

• Data Labeling: Supports manual labeling (e.g., for Label Studio) and weak supervision techniques using Snorkel.
• Data Augmentation: Implements various augmentation strategies for text, images, and tabular data (e.g., SMOTE).
• Imbalanced Data Handling: Provides methods like class weighting and focal loss to address skewed datasets.
• Data Splitting: Offers robust splitting strategies including random, temporal, and group splits to prevent data leakage.
• Use Case: You have an imbalanced dataset for a fraud detection model. Use this Skill to apply SMOTE for oversampling the minority class and then perform a stratified split to ensure both training and testing sets accurately represent the class distribution.

Quick Start

Use the training-data skill to perform a stratified split on your dataset with a 80/20 train/test ratio.