Machine Learning Thursdays: Replacing the Tractor, Not the Farmer, with Predictive Analytics
There’s a concern that pops into the heads of users when they see the power of automation through SAP Predictive Analytics. It sounds very similar to how people were unsure of accepting the industrial revolution as the next best thing for all since they feared losing their livelihood. This trend continued when low skilled workers like stokers (firemen) had to learn something else to be relevant when steam engines became outdated. The more powerful engines replacing the steam engine actually benefited those same people by improving their travel experience, which became faster and cleaner.
When we talk about complex, advanced topics like predictive analytics and machine learning, we don’t see the likes of firemen in the picture in the first place. The power of automation that we bring with our solution doesn’t cause redundancies—it actually helps key resources in business like business analysts and citizen data scientists make the best of their valuable time by taking care of mundane tasks. They can focus on the bigger picture of how they can generate robust and reliable insights that’ll help their company have an edge over competition.
Consider how a farmer with a better tractor can manage a larger piece of land and produce more efficiently, but a tractor still needs a farmer to operate and maintain it. Similarly, SAP Predictive Analytics helps managing predictive tasks better, but the job still has to be done by the business analyst or citizen data scientists.
So How Is SAP Predictive Analytics the Better Tractor?
SAP Predictive Analytics is a suite of solutions that help users through every phase of predictive analytics. Once you have your business problem identified, you can prepare powerful analytical datasets with Data Manager, build robust models with automated or expert analytics, and automate the management of a predictive model’s lifecycle.
Prepare Powerful Analytical Datasets with Data Manager
- Build analytical datasets with clicks; no coding necessary.
- Works with all kinds of data (geographical, times series, …)
- Create thousands of derived features/columns to build robust models (again without coding)
- Identify which variables are changing over time with timestamped populations
- Generate automatically reusable SQL code and documentation
Build Robust Models with Automated or Expert Analytics
The suite provides two powerful options for building predictive models: Automated Analytics and Expert Analytics. The two options combined offer a host of powerful capabilities to:
- Leverage SAP’s patented automated functions to build robust models easily that does more than just train an algorithm, automatically
- Leverage the power of SAP HANA not just as a data source but as a predictive engine by using in-database libraries like Automated Predictive Library (APL) and Predictive Analytics Library (PAL)
- Extend functionality of the solution by bringing your own R (BYOR) scripts and pipe it in with the in-database libraries on SAP HANA
- Export models into SAP HANA so predictive models can be directly embedded into business process/applications
Automate Management of Predictive Model’s Lifecycle
- Schedule recurrent model training and application on new data
- Automatically detect data deviation and trigger retraining of obsolete models
- Time and Event based scheduling of both retrain and apply tasks
- Monitor quality of thousands of generated models
How Does All This Benefit the Business Analyst, Citizen Data Scientist, and Most Importantly, the Business?
- Accurate results in days, not weeks: Accelerate your work with automated techniques in Data Preparation, Modelling and Model Management, without sacrificing performance and quality
- Machine learning at scale: Maintain peak performance for hundreds or more models through massive automation
- Insights where people interact: Embed predictive models in business processes and applications, especially SAP business applications such as SAP HANA, SAP Cloud and SAP Fraud Management