Machine Learning Thursdays: Why Automation Is Critical in Making Machine Learning Pervasive in Businesses
First, we need to look at how Machine Learning can become pervasive in business.
1)One way is to make the creation cost of predictive models lower by reducing the level of effort and skill needed to build them so that citizen data scientists can be effective.
- Today, we do this with SAP BusinessObjects Predictive Analytics Automated mode.
- We make it easy to compare models, we make it highly robust across differing data types, and we automatically produce analytical data sets and scoring equations to apply models to new data.
But this will only take you so far. A user needs to be able to formulate the business question correctly using the software and needs to find the data needed to answer it.
2)Another way is to build marketplaces of machine learning IP where a vendor can solve a problem and resell the solution to end customers.
- This reduces the data science effort but the end customer still needs to acquire the solution and needs time and skill to be able to evaluate it and implement it.
3)The final way is to embed the model directly in the business application where a business user can see the results of predictive modelling without needing to be a data scientist.
- The SAP Predictive Analytics team is working with application teams such as SAP S4/HANA to do just this and we have already shipped SAP Fraud Management with embedded predictive models completely built in. Soon we’ll ship machine learning in SAP S4/HANA applications, which repeat the pattern.
- There are thousands of use cases out there which could help drive real business value for customers, so we plan to partner with organizations like Accenture to be able to scale out the number of solutions we support using partner-developed solutions.
“By 2018, 75% of enterprise and ISV development will include cognitive/AI or machine learning functionality in at least one application, including all business analytics tools.” IDC
Automated Machine Learning
Given this, where does automated machine learning come in and why is it important? First, applications are built to support a wide variety of business models and usage. They vary in the fields they add, the number of records they store, and the types of business operations they need to manage.
Automated Modeller has capabilities that enable it to be highly robust across very varied data profiles and to generate SQL-based scoring equations which automatically adjust if the model needs to adjust. This means that we can ship features that work well out of the box regardless of the customer.
We also know that customers don’t want black boxes, so we enable data scientists to access debrief information so they get all the information they need to help them understand that it’s robust and accurate. And if they wish to, they can build a new model and replace the model we ship out of the box with something that is more adapted to their environment.
We are currently ramping up the number of use cases and applications we support, which means that the best way to make machine learning pervasive in your business is to be an SAP Applications and Predictive Analytics customer.
Because of the investments we make in automated modelling, you can be confident these models will work well regardless of your company size or usage. And if you wish to improve them we will have the technology and partners to help you achieve this.
- In this report, Forrester shares seven attributes to look for when selecting a predictive analytics and machine learning solution. Forrester Wave on Predictive Analtyics and Machine Learning
- Access this recent TDWI report, “Machine Learning for Business: Eight Best Practices for Getting Started,“ for an overview of key activities companies should engage in to drive machine learning adoption. TDWI Paper on Predictive Analtyics and Machine Learning
- Read our other blogs on machine learning and predictive analytics topics.