A Visual Introduction to Machine Learning - Vlad Golubev

A Visual Introduction to Machine Learning - Vlad Golubev

Transcript

1. A Visual Introduction to Machine Learning Specially for GDG Ternopil
2. Vladyslav Golubev linkedin.com/in/vladgolubev github.com/vladgolubev golubev.vld@gmail.com Who am I Meteor Machine Learning Node.js Microservices 1 year!
3. Why it is important? • Trend: 90% of the data in the world today has been created in the last two years • Recommendation systems: Amazon, eBay, IMDb • Text & Speech recognition: Google • Sales forecasting: Everywhere Use cases: • Problem: people can’t operate on such amounts • Solution: apply machine learning to provide valuable insights Predictions, not causality
4. Machine Learning
5. — What the heck is Machine Learning anyway? Definition Machine Learning is a field of study that gives computers the ability to learn without being explicitly programmed.
6. — What the heck is Machine Learning anyway? Definition Machine Learning is a field of study that gives computers the ability to learn without being explicitly programmed. … explores algorithms that can learn from … … and make prediction on data
7. size color shape fruit big red round apple small red heart-shaped cherry big green long cylinder banana small green oval grape Okay, we have some data features decision variable How does it look like? observations usually muuuch more wider
8. Basic concepts: Train vs Test data Train data: • outcome is known • used for learning Test data: • outcome is unknown • used for classifying size color shape fruit big red round apple small red heart-shaped cherry big green long cylinder banana small green oval grape size color shape fruit big red round ? small red heart-shaped ? big green long cylinder ? small green oval ?
9. Basic concepts: Supervised vs Unsupervised Supervised: • Usually classification • 70% of machine learning Unsupervised: • Usually clusterization • Group similar data input
10. Basic concepts: Unsupervised learning No outcome in train data! apple cherry banana grape
11. Basic concepts: Supervised learning size color shape fruit big red round apple small red heart-shaped cherry big green long cylinder banana small green oval grape Train data: Is size big? apple cherry Is shape oval? Is color red? yes no grape banana Decision trees!
12. Basic concepts: Decision trees Assuming we have this dataset elevation year bathroom bedroom price area city 120m 1950 2 1 514500 80 San Francisco 21m 1982 1 2 357800 45 New York 198m 2000 3 3 1100000 120 San Francisco 15m 1993 4 5 850000 30 New York …. …. …. …. …. …. ….
13. Basic concepts: Decision trees It’s all about finding patterns 3. Repeat 1. Look at one variable at a time 2. Use statistical learning to identify boundaries best split elevation greedy
14. Decision trees Training 1st split gives decent result zero benchmarks
15. Testing • overfitting • crossvalidation Decision trees
16. Real-life example: Tools
17. Real-life example: Kaggle Titanic competition Loading the datasets
18. Real-life example: Kaggle Titanic competition First data investigation
19. Real-life example: Kaggle Titanic competition Adding custom features
20. Real-life example: Kaggle Titanic competition Creating a decision tree and predicting!
21. Useful Resources • Datacamp.com - the easiest way to learn data science • Kaggle.com - the home of data science • https://www.coursera.org/learn/machine-learning/
22. Many thanks. Vladyslav Golubev linkedin.com/in/vladgolubev github.com/vladgolubev golubev.vld@gmail.com