Like no other technology, Artificial Intelligence has the ability to incite both intense fear and optimistic hype – often in the same breath. There is no shortage of examples in popular culture where on the one hand, intelligent machines lead to apocalyptic scenarios where humans are subjugated to organic conscious waste, and on the other where Artificial Intelligence (AI) leads to a utopic society where machines do all thinking and hard work for us – leaving plenty of time for us to sit back and enjoy life’s simple pleasures.
The reality is far from both extremes. AI and Machine Learning (ML) have both come on leaps and bounds in recent years, leading to many innovative applications of the technology. The technology is still nascent and largely based on advanced statistical methods and mathematics at the basic level. However, the technology is allowing for complexity through modelling neural networks that mimic elements of human cognitive function.
AI & ML help create the intelligent enterprise
Placing AI and ML into some kind of enterprise context is an immense challenge for many organisations. The technology offers the potential for huge competitive advantage to those organisations that integrate the latest and greatest machine learning and deep learning models into their enterprise. This approach will help to solve real business problems, give advanced insight, prescribe action on this insight and provide a level of automation to specific skills and business processes.
Keytree brings a raft of experience in working with machine learning, bringing in our culture of innovation and design-led thinking to build advanced analytical applications that are powerful but yet are highly visual, intuitive and easy to use. Some recent examples where ML and Advanced Analytics has been transformative include:
A sports injury prevention solution built in partnership with leading sports scientists where Keytree developed custom ML algorithms to provide advanced insight into the nature of sports injury. Acting as an aid to injury prevention, and also as a research tool to better understand serious injury in sport.
Keytree In-store Technology (KIT) enhances the digital and in-store experience for staff and customers by providing a unified view of customer interactions, purchasing history, product, and stock availability. KIT empowers store associates and retail managers by providing required customer and stock information in-store at the push of a button, improving efficiency, productivity and, for the customer, the overall shopping experience.
An HR Analytics solution, which leverages SAP’s native ML to help better understand the relationships that map people to performance, help measure and control these drivers. The solution supports planning and advanced scenario analysis allowing for actions to be taken – aimed at improving and driving better people and financial performance results.
Navigating through the various SAP tooling and offerings related to AI, ML and predictive analytics is challenging. This blog attempts to explain and bring together the SAP offerings in this space both on-premise and in the cloud.
SAP Predictive Analytics & Machine Learning
SAP has a broad offering for applying ML into an SAP landscape, and SAP HANA is at the heart of enabling this advanced ML capability. SAP HANA provides the processing power required to process complex mathematical algorithms by leveraging in-memory processing and massive parallel processing (MPP) technology to provide the necessary speed and scale.
For ML, SAP has various tooling options and these tools provide various levels of abstraction to the underlying complexity of ML algorithms – depending on whether you are an end user or a developer/data scientist.
The Predictive Analysis Library (PAL) – At the core of HANA’s ML offering is an extensive set of in-memory algorithms available by default in a HANA installation.These algorithms can be accessed via SQL script, graphically modelled inside HANA in flow-graphs or used to deliver various predictive or ML applications.The PAL provides in excess of 100 algorithms providing advanced ML models and capabilities which include the following areas:
- Association – to build intelligent associates, typically used in recommendation and basket analysis
- Classification – to predict who will and what is the probability of certain behaviour e.g. who will buy, churn next week, month or year
- Regression – to predict what will be e.g. revenue, the number of churners, next week, month or year
- Clustering – to Identify clusters of groups with similar behaviour and profiles, and help identify outliers
- Forecasting – to build predictions on what will be e.g. revenue, the number of churners next week, month or year
Automated Predictive Library (APL) – The APL is aimed at automating several steps of ML. Among other things, it generates and picks the best model to perform the ML function and automates the workflow around usage. Although it does run on HANA, it is not part of the standard HANA license.
Application Function Library (AFL) – The AFL allows third-party build of external libraries that are linked to the SAP HANA database process during runtime, and are highly optimised and offer high performance. AFL’s contain application functions that are implemented in C++ and that can be started by executing special database procedures. The PAL forms part of the APL.
R-Server – R is an open source programming language for statistical computing and graphics. The R language is widely used among statisticians and data miners for developing ML software and data analysis. HANA can run R-Scripts on an R-Server. Procedures are created and reside in SAP HANA, but the execution is handled by an R-server, which is on separate hardware.
External Machine Learning Library (EML) – TensorFlow is an open source software library for high-performance numerical computation. Originally developed by researchers and engineers from the Google Brain team within Google’s AI organisation, it comes with strong support for Machine Learning and deep learning for building complex neural networks modelled on the human brain and nervous system.
The flexible numerical computation core of TensorFlow can be used across many other scientific domains. HANA 2.0 SPS02 includes the External Machine Learning library which makes this possible. The EML is packaged as an Application Function Library (AFL). Models served by TensorFlow are registered in HANA via remote sources and then accessed through SQL-script.
Come back next week for part two where I will discuss the various cloud-based offerings available on the SAP Cloud Platform, SAP Analytics Cloud and also available as part of the Leonardo Machine Learning foundation.