R-_SAP_Hana.pdf

Transcript

1. Using R for machine learning with SAP HANA Dr Maria Prokofieva Business School, Victoria university, Australia R Ladies Melbourne
2. Data Science
3. Data Science • Business decisions driven on data insights • Big data • Artificial intelligence • Visualisation
4. What is R? - Started by Ross Ihaka and Robert Gentleman in 1995 - GNU project, based on S language - Free - Open source programming language and programming environment (Rstudio) - Powerful and highly extensible coding tool.
5. Why R? • The easy-to-understand syntax • RStudio • R packages • Data handling and large data sets • Supportive community
6. Why R? Source Which languages are important for Data Scientists in 2019?
7. Why R and SAP HANA • Embedding of R code in the SAP HANA database • R script is executed as part of the database query • No additional data copy in R environment
8. R and SAP HANA integration • R/Rstudio www.r-project.org • Rserve: www.rforge.net/Rserve • TCP/IP server to use R from various languages • No need to initialize R or link against R library. • separate workspace and working directory for every connection
9. R and SAP HANA integration R code
10. Machine Learning • “All models are wrong but some are actually useful” (G. Box) • ML vs artificial intelligence • ML vs statistics • Supervised vs Unsupervised algorithms • Deep learning • Augmented analytics
11. ML Tasks • Classification • Dimension reduction • Clustering • Neural networks
12. R packages and CRAN • network of ftp and web servers around the world that store R code and documentation • ML repository
13. Caret Package • Classification and REgression Training (CARET) package. • General package for creating ML workflow to combine model training and prediction • Allows to run several different algorithms for a given business problem to estimate the most accurate/efficient one • names(getModelInfo()) : 217 possible methods to be used in a single model • To build a predictive model: train(formula, data, method)
14. MICE Package • Managing missing values • impute missing values by using multiple techniques • complete(data, action, ...)
15. Specific ML packages • randomForest: random forests • Glmnet: regularized regression models • e1071: Support Vector Machine • rpart: classification and regression trees
16. Deep Learning: TensorFlow and Keras • algorithms inspired by the structure and function of the brain = artificial neural networks • free open-source software library for dataflow • developed by Google Brain Team – Machine Learning Crash Course Keras • high level API built on TensorFlow • “Being able to go from idea to result with the least possible delay is key to doing good research”.
17. Business cases • Text analytics • Social media analytics • Image analytics