H2O World - Intro to Data Science with Erin Ledell
7 likes3,031 views
This document provides an introduction to data science. It defines data science as using data to solve problems through the scientific method. The roles of data scientists, data analysts, and data engineers on a data science team are discussed. Popular tools for data science include Python, R, and APIs that connect data processing engines. Machine learning algorithms are used to perform tasks like classification, regression, and clustering by learning from data rather than being explicitly programmed. Deep learning and ensemble methods are also introduced. Resources for learning more about data science and machine learning are provided.
H2O World - Intro to Data Science with Erin Ledell
- 1. Intro to Data Science Erin LeDell Ph.D.โจ Statistician & Machine Learning Scientistโจ H2O.ai
- 2. H2O World 2015 Download our app, โH2O World 2015โ
- 3. H2O World 2015 I have H2O Installed I have Python installed I have R installed I have the H2O World data sets Pick up stickers or get install help at the information booth
- 4. Intro to Data Science โข What is Data Science? โข The Data Scientist โข The Data Science Team โข Data Science Tools โข What is Machine Learning? โข What is Deep Learning? โข What is Ensemble Learning? โข Data Science Resources
- 5. What is Data Science? One of the earliest uses of the term "data science" occurred in the title of the 1996 International Federation of Classification Societies conference in Kobe, Japan.
- 6. What is Data Science? โข The term re-emerged and became popularized in 2001 by William Cleveland, then at Bell Labs, when he published, โจ "Data Science: An Action Plan for Expanding the Technical Areas of the Field of Statisticsโ. โข This publication describes a plan to enlarge the major areas of technical work of the field of statistics. Dr. Cleveland states, "Since plan is ambitious and implies substantial change, the altered field will be called Data Science."
- 7. What is Data Science?
- 8. What is Data Science? โข Clean, transform, filter, aggregate, impute โข Convert into X and Y Problem Formulation Data Processing Machine Learning โข Identify a data task or prediction problem โข Collect relevant data โข Train models โข Evaluate models
- 9. The Data Science Venn Diagram Drew Conway (2010)
- 11. The Data Scientist โUnicornโ
- 12. Survey of Data Scientists on LinkedIn The number of data scientists has doubled over the last 4 years. The top five skills listed by data scientists: 1. Data Analysisโจ 2. R 3. Python 4. Data Miningโจ 5. Machine Learning
- 13. From Data Unicorns to Data Teams
- 14. Data Science Teams โข Usually a background in computer science or engineering โข Very good programming and DevOps skills Data Analysts Data Engineers Dataโจ Scientists โข Strong data skills and the ability to use existing data analysis tools โข Able to communicate and tell a story using data โข Strong math/stats background in addition to programming ability โข Understanding of machine learning algorithms
- 16. Data Science in the Enterprise โข Data Science teams develop โactionable insightsโ for business. โข They provide decision makers with information, guidance and confidence in the decision making process. โข Competitive advantage โข Cost minimization โข Data-driven products
- 17. Data Science in the Enterprise Donโt be a data dinosaur. Embrace the data!
- 18. Data Science Tools 2013 was the year of the โจ data science โlanguage wars.โ
- 19. Data Science Tools In 2015, we have evolved beyond thisโฆโจ We are too busy doing actual data science!
- 20. Data Science Tools We are headed toward language agnostic data science, where friendly APIs connect to powerful data processing engines.
- 21. What is Machine Learning? Unlike rules-based systems which require a human expert to hard-code domain knowledge directly into the system, a machine learning algorithm learns how to make decisions from the data alone. โField of study that gives computers the ability to learn without being explicitly programmed.โ โจ โ Arthur Samuel, 1959
- 22. Machine Learning Tasks โข Multi-class or binary classification โข Ranking (e.g. Google Search results order) โข Evaluate with Classification Error or AUC Regression Classification Clustering โข Unsupervised learning (no training labels) โข Partition the data; identify clusters or sub-populations โข Evaluate with AIC, BIC or Total Sum of Squares โข Predict a real-valued response (e.g. viral load, price) โข Gaussian, Gamma, Poisson, etc. distributed response โข Evaluate with MSE or R^2
- 23. Train, Validation and Test Set โข If you plan on doing any model tuning, you should split your dataset into three parts: Train, Validation and Test โข There is no general rule for how you should partition the data and it will depend on how strong the signal in your data is, but an example could be: โจ 50% Train, 25% Validation and 25% Test โข The validation set is used strictly for model tuning (via validation of models with different parameters) and the test set is used to make a final estimate of the generalization
- 24. K-fold Cross-validation โข K-fold Cross-validation (CV) โจ is used to evaluate the performance of machine learning algorithms. โข CV will give you the most โmileageโ on your training data. โข Performance metrics are averaged across k folds.
- 25. Machine Learning Workflow Training and Predictionโจ in machine learning
- 26. What is Deep Learning? โข Deep neural networks have more than one hidden layer in their architecture. Thatโs why they are called โdeepโ neural networks. โข Very useful for complex input data such as images, video, audio. โA branch of machine learning based on a set of algorithms that attempt to model high-level abstractions in data by using model architectures, composed of multiple non-linear transformations.โ โจ โจ โ Wikipedia (2015)
- 27. What is Deep Learning? โข Deep learning architectures, specifically artificial neural networks (ANNs) have been around since 1980. โข However, there were breakthroughs in training techniques that lead to their recent resurgence in the mid 2000โs. โข Combined with modern computing power, they are quite effective.
- 28. What is Ensemble Learning? โข Random Forests and Gradient Boosting Machines (GBM) are both ensembles of decision trees. โข Stacking, or Super Learning, is technique for combining various learners into a single, powerful learner using a second-level metalearning algorithm. โEnsemble methods use multiple learning algorithms to obtain better predictive performance that could be obtained from โจ any of the constituent learning algorithms.โ โจ โ Wikipedia (2015)
- 29. No Free Lunch โข No general purpose algorithm to solve all problems. โข No right answer on optimal data preparation. โข Some algorithms may have such strong biases that they can only learn certain kinds of functions. "Even after the observation of the frequent or constant conjunction of objects, we have no reason to draw any inference concerning any object beyond those of which we have had experience.โ โจ โ David Hume (1711-1776)
- 30. Where to Learn More? โข H2O Online Training (free): http://learn.h2o.ai โข H2O Slidedecks: http://www.slideshare.net/0xdata โข H2O Video Presentations: https://www.youtube.com/user/0xdata โข H2O Community Events & Meetups: http://h2o.ai/events โข Machine Learning & Data Science courses: http://coursebuffet.com