HANA 2.0 SPS 01 was recently released. If you want to learn about and see working examples of what’s new for the predictive analysis library (PAL) please read on.
A number of important new predictive and machine learning algorithms have been added to the PAL as well as some key enhancements to existing algorithms.
Real time scoring is now possible for selected algorithms.The aim is to optimize performance and response times when predictions from complex models are required frequently. With the PAL, once a model has been trained it's stored in a HANA table encoded in XML or JSON or PMML. Before the model can be used to make predictions it needs to be
un-encoded into memory. Typically this is very fast but for some algorithms the models can be large leading to the un-encoding process requiring more time than actually making a prediction!. Support Vector Machine and Random Forest are good examples of this. To better support scenarios where predictions need to be made frequently (i.e. in real time!) - as part of a transaction or perhaps against streaming data - it's now possible to explicitly load the model into memory in advance (i.e. create a "state") so the
un-encoding is only performed once allowing individual calls to make their predictions blindingly fast.
Recommender systems is an interesting new category that exploits state-of-the-art machine learning algorithms to predict (or recommend!) ratings or preferences. In our example we use the Factorized Polynomial Regression Models algorithm to analyze ratings provided by SAP HANA Academy team members for the all-time Top 30 best-selling albums. We then use this model to predict ratings of team members for other albums not included in the all-time Top 30. One of the key features of Factorized Polynomial Regression Models is it's ability to make good use of
side features - extra information about an item, user, or relevant to when the rating was made. For an album it might be the artist or style of music, for a user it could be their gender, or it could be the user's location when a making a rating.
To help get started with all these new capabilities the SAP HANA Academy has created hands-on video tutorials. In case you're new to SAP HANA and the predictive analysis library, the getting started tutorial has been extended to include an full overview of predictive with SAP HANA - so there's no need for prior knowledge in order to get going.
Here are direct links to the hands-on video tutorials:
Getting Started
Intro and Overview
Real Time Scoring
Create Model State
Predict with Model State
Delete Model State
Statistics
One-Sample Median Test
Wilcox Signed Rank Test
TTest
ANOVA One-Way
ANOVA One-Way Repeated Measures
Clustering
Accelerated Kmeans
Classification
Support Vector Machine (SVM) Outlier Detection - Train
Support Vector Machine (SVM) Outlier Detection - Predict
Recommender Systems
Factorized Polynomial Regression Models - Train
Factorized Polynomial Regression Models - Predict
Here’s the full playlist for what’s new in predictive in this release:
What’s New for Predictive HANA 2.0 SPS01.
One final point, all the video tutorials were recorded on
SAP HANA, express edition on the Google Cloud Platform - so you can easily try them yourself and take advantage of the
GCP free trial.
If you’re interested to learn about what's new with HANA 2 SPS 01 in general check out the following
blog.
Happy machine learning with HANA!
-
The
SAP HANA Academy provides free online video tutorials for the developers, consultants, partners and customers of SAP HANA.
Topics range from practical how-to instructions on administration, data loading and modeling, and integration with other SAP solutions, to more conceptual projects to help build out new solutions using mobile applications or predictive analysis.
For the full library, see
SAP HANA Academy Library - by the SAP HANA Academy
For the full list of blogs, see
Blog Posts – by the SAP HANA Academy
