Long recognized as possessing strong cloud capabilities—virtualization, shared resources, EAL5+ security ratings and more—z/OS is the preferred choice for many of the top banks, retailers and insurance organizations in the world, who process billions of transactions per day
through their mainframes, handling some of the world’s most business critical and valued data. All of this activity, when combined with rich volumes of new and historical transactions, creates a prime opportunity to exploit cognitive capabilities such as machine learning to help produce smarter business outcomes.
From Dumb to Smart Applications
Just as green screens were referred to as “dumb terminals” in the 80s and 90s, there’s a new breed of application exploiting machine learning that makes applications absent of the technology look like “dumb applications.”
Think of the applications that run on z Systems. These are often mission-critical and involve vast sums of money, not only at the transaction level, but also in the associated risks, opportunities or missed opportunities if trends and patterns are not immediately identified. Equally bad would be fraud that remains undetected and is repeatedly exploited. Machine learning
has the potential to be the new app that could help businesses get continually smarter, allowing them to learn from new data at each interaction. At the same time, IT departments struggle to keep up with the rapid pace of business change — not just in speed to market of writing new applications, but in maintaining existing apps that contain increasingly complex business rules and logic.
For all of these reasons and more, IBM is delivering the full range of our machine learning capabilities to z/OS with IBM Machine Learning for z/OS
, essentially bringing advanced machine learning to the world’s most valued data. Welcome to the learning machine.
Not All Platforms Are Born Equal
Any platform that handles thousands of transactions per second will always be an attractive business computing environment. However, the personal and high value data stored on such platforms can become very attractive to cyber criminals and thieves as the potential financial gains if breached far outweigh the risks of being caught. IBM z Systems’ security capabilities and EAL5+ certification can help combat such threats. Combine these security features with the platform’s processing power and scalability and you have an ideal environment to deploy and exploit machine learning to help rapidly identify emerging patterns, predict and help prevent potential fraud, and identify opportunities faster and more frequently than a human ever could.
Below are some of the key capabilities of machine learning from IBM.
Whether the user is an experienced data scientist or a beginner, Machine Learning for z/OS offers a unified set of powerful capabilities through a variety of interfaces for interaction—from Jupyter notebooks to a more intuitive graphical interface. There’s also support for Python, Java and Scala to allow organizations to leverage their preferred programming language and skills when building machine learning applications. Organizations can also deploy Machine learning from IBM across different computing environments from private cloud to public cloud—including IBM z Systems z/OS and choose from frameworks such as SparkML, TensorFlow and H20.
With the data available for machine learning solutions, users can create advanced algorithms or choose from a set of powerful, predefined algorithms and models without requiring advanced data science expertise.
IBM Machine Learning was built around three core principles: Simplicity, collaboration (across multiple personas) and convergence of a wide range of technologies from the IBM analytics portfolios and IBM research laboratories. The user experience is key. Across personas, IBM Machine Learning helps users engage and collaborate on machine learning projects, leveraging the combined skills of the team. Wizards within the tools provide step-by-step processes and workflows that automate many aspects of building, testing and deploying true learning machines—those that deploy cognitive applications and are able to get smarter with each interaction. With features such as IBM Cognitive Assistance for Data Scientists, the system chooses the best algorithm for the user based on a training data set.
Once a model is built and tested, it needs to be deployed. A model—in fact the entire machine learning application—is similar to a living organism, evolving and adapting over time as it encounters new data with each transaction and interaction. A continuous feedback loop enables the model to adapt and change, altering specific parameters within the model itself in order to become smarter and more consistent over time—while avoiding what is known as “overfitting,” or becoming too accurate to handle any deviation from its expected outcome. This auto-tuning reduces manual intervention. Even so, some human intervention or model adaptation may be necessary where a human judgment or decision is required. Therefore, keeping track of the version of the models over the lifecycle of the learning machine is important for audit purposes or for falling back to a previous version.
These capabilities can be visualized in Figure 1 below.
Figure 1: IBM Machine Learning
The IBM Machine Learning Hub
Beyond machine learning technology, IBM recognizes the need to nurture and collaborate with organizations as they embrace and fully exploit its machine learning technologies. The first IBM Machine Learning Hub
is designed to help accelerate and enrich organizations’ machine learning knowledge and expertise through access to IBM world-class data science professionals who can provide education and training, expert advice on all aspects of machine learning — as well as lead and deliver proofs-of-concept and full client engagements. The combination of the technology aspects and the knowledge/skills base provides an opportunity for a unique machine learning experience.
Watch a short video on machine learning
and read the recent announcement of IBM Brings Machine Learning to the Private Cloud
There are also machine learning courses available, but it can be hard to choose which course is right for you. To learn more about available courses, visit "The Best Machine Learning Courses for All Levels."
Dinesh Nirmal is vice president of Analytics Development. He is on Twitter at @DineshNirmalIBM.