How does it help prevent data leakage?

It enforces strict separation of training and testing data, ensuring that transformers like Scalers are only fitted on the training set before being applied to validation and test sets.

Can I use this for deep learning and neural networks?

Absolutely. It includes best practices for neural network architecture, including dropout for regularization, batch normalization, and early stopping to prevent overfitting.

Which machine learning frameworks does this skill support?

The skill provides comprehensive support and implementation patterns for scikit-learn, PyTorch, and TensorFlow/Keras.

Does it handle imbalanced datasets?

Yes, it includes specific patterns for using class weights, SMOTE oversampling, and balanced evaluation metrics to handle minority classes effectively.

Does it help with model reproducibility?

Yes, it provides a standardized way to set random seeds across Python, NumPy, and PyTorch to ensure your experiments are consistent and reproducible.

Machine Learning Model Training

Name: Machine Learning Model Training
Author: secondsky

bysecondsky

•

データサイエンスとML

Trains production-ready machine learning models using industry-standard frameworks like scikit-learn, PyTorch, and TensorFlow.

概要

This skill empowers Claude to handle the entire machine learning lifecycle, from robust data preprocessing and feature engineering to model selection and evaluation. It provides standardized implementation patterns for supervised learning tasks, deep neural networks, and hyperparameter tuning while actively preventing common pitfalls like data leakage, class imbalance, and overfitting. Whether you are building a simple scikit-learn classifier or a complex PyTorch training loop with early stopping, this skill ensures best practices are followed for reproducible and reliable AI models.

主な機能

Built-in prevention strategies for data leakage and class imbalance
21 GitHub stars
Comprehensive evaluation metrics for both classification and regression tasks
Automated data preparation including scaling, encoding, and stratified splitting
Multi-framework support for scikit-learn, PyTorch, and TensorFlow/Keras
Advanced training loops with early stopping, checkpointing, and regularization

ユースケース

Implementing deep learning architectures with proper dropout and normalization
Building and tuning predictive models for business classification or regression
Debugging model performance issues like convergence failure or overfitting

概要

主な機能

Built-in prevention strategies for data leakage and class imbalance
21 GitHub stars
Comprehensive evaluation metrics for both classification and regression tasks
Automated data preparation including scaling, encoding, and stratified splitting
Multi-framework support for scikit-learn, PyTorch, and TensorFlow/Keras
Advanced training loops with early stopping, checkpointing, and regularization

ユースケース

Implementing deep learning architectures with proper dropout and normalization
Building and tuning predictive models for business classification or regression
Debugging model performance issues like convergence failure or overfitting