应用机器学习:特征工程

应用机器学习:特征工程

The quality of the predictions coming out of your machine learning model is a direct reflection of the data you feed it during training. Feature engineering helps you extract every last bit of value out of data. This course provides the tools to take a data set, tease out the signal, and throw out the noise in order to optimize your models. The concepts generalize to nearly any kind of machine learning algorithm. Instructor Derek Jedamski provides a refresher on machine learning basics and a thorough introduction to feature engineering. He explores continuous and categorical features and shows how to clean, normalize, and alter them. Learn how to address missing values, remove outliers, transform data, create indicators, and convert features. In the final chapters, Derek explains how to prepare features for modeling and provides four variations for comparison, so you can evaluate the impact of cleaning, transforming, and creating features through the lens of model performance.

Topics include:

  • What is feature engineering?
  • Exploring the data
  • Plotting features
  • Cleaning existing features
  • Creating new features
  • Standardizing features
  • Comparing the impacts on model performance

课程信息

  • 英文名称:Applied Machine Learning: Feature Engineering
  • 时长:2小时26分
  • 字幕:英语

课程目录

  1. The secret of effective machine learning
  2. What you should know
  3. What tools you need
  4. Using the exercise files
  5. What is machine learning?
  6. What does machine learning look like in real life?
  7. What does an end-to-end machine learning pipeline look like?
  8. What is feature engineering?
  9. Why does feature engineering matter?
  10. What are the tools in the feature engineering toolbox?
  11. What data are you using?
  12. Explore continuous features
  13. Plot continuous features
  14. Explore categorical features
  15. Plot categorical features
  16. Summary of features
  17. Treat missing values in the data
  18. Cap and floor data to remove outliers
  19. Transform skewed features
  20. Creating new features from text
  21. Create indicators
  22. Combining existing features into a new feature
  23. Convert categorical features to numeric
  24. Create training and test sets
  25. Standardize all features
  26. Write out three final datasets
  27. Review model evaluation basics
  28. Build a model with raw original features
  29. Build a model with cleaned original features
  30. Build a model with all features
  31. Build a model with reduced set of features
  32. Compare and evaluate all model variations
  33. How to continue advancing your skills

评论