Matthew Roberts designs process improvement solutions for IBM clients across the UK and Ireland. As a member of the UKI technical sales team, Matt leads the adoption of IBM Operational Decision Manager Advanced within systems of insight. In this role, Matt builds proof of concept prototypes and leads the technical specification of new Smarter Process solutions. Matt is a certified developer for IBM Operational Decision Manager and IBM Business Process Manager. He holds a Master of Computer Science degree from the University of Oxford.
Systems of insight are ever analyzing and monitoring what is going on in the world, seeking out new opportunities and identifying situations that need attention and action. These systems provide the ability to apply analytics and rules to real-time data as it flows within and beyond the enterprise to gain the desired insight. Deriving this insight is a key step towards being able to make the right decisions and take the correct actions.
Systems of insight are found in counter-fraud management, promotions and pricing solutions, travel and transportation scenarios, and many other domains. The common characteristic is that each system requires the detection of a situation (for example, a pattern of customer behavior) and determination of the next-best-action.
IBM® Operational Decision Manager Advanced (ODM), provides a way to deliver these use cases through enterprise-scale detect and decide automation capabilities. This is achieved through a combination of event-driven situation detection and request-driven decision automation.
In this article, we outline some of the systems of insight usage patterns that include ODM, as introduced in chapter 2 of this IBM Redbooks publication, Systems of Insight for the Digital Transformation Era, SG248294.
Event-driven situation detection
The simplest pattern occurs where ODM directly consumes data from event sources. Event sources are often batch events and transactional feeds from systems of record, streams data from sensors or application and process events from systems of engagement including mobile devices and business applications. After detection and decision logic has been applied within ODM, the emitted events are used to trigger actions. Commonly the action takes the form of a notification to a user or the invocation of a business process.
This pattern can be applied to detect payment fraud. For example, the situation detection system (ODM) correlates events that represent different transactions to spot suspicious behavior. Flagged transactions are stopped for investigation by a case handler.
Andy Ritchie and Dan Selman explore the architecture of event-driven situation detection solutions in their article, Event Driven Architecture and Decision Management.
Stream-based situation detection
In some circumstance the structure of data being received by a system of insight isn't appropriate for direct consumption by ODM. For example, a video-feed, or high-volume and low-value sensor event streams require can require specialist stream-processing analytics to convert the raw data into a form that is useful for making a business decision.
Take a social-media feed for example, understanding the sentiment and content of large volumes of natural language can be achieved through solutions such as the IBM Insights for Twitter service on Bluemix. When a particular tweet has been analyzed we can then use ODM to either decide what to do next, or respond to a pattern of tweets about a particular product or competitor for example.
The two diagrams in this section show different integration patterns for stream processing with ODM. In the first, situation detection sits after stream processing in the pipeline and works on the higher-value events emitted by the stream processing solution. In the second diagram, a decision service is used as part of the stream analytics to help it detect patterns directly. This architecture is used in the IBM Next Best Action solution.
Analysis-driven decision automation
Alternatively, where insight is gained within analysis of static data, analytical modelling tools can be used first to help determine what the decision automation logic should be. This occurs through the collection and subsequent analysis of data which then informs the structure and thresholds of business rules that automate future decisions. The business rules are encapsulated through an exposed decision service that provides an interface for use by different business applications. Calling applications then request a decision in context of the application's flow, for example when a user presses a button or during a transition between web-pages. This might be to calculate the price for a hotel booking where promotion logic was initially driven through analysis of customer demand and competitors' offers.
Typically this initial manual analytic investigation is performed using tools such as IBM i2 Intelligence Analysis Platform or IBM SPSS Modeller.
Big data decision automation
Finally, in large-scale distributed processing of big data the convention is to use scripting or programming languages to implement processing logic. For example, on a Hadoop platform, Java code applied in a MapReduce framework allows highly distributed processing; alternatively, query languages such as Big SQL in the IBM InfoSphere BigInsights platform make these large data stores more accessible to IT professionals with existing SQL skills. In both of these examples the result of execution provides an answer that can be used to make a decision.
By combining Big Data with Business Rules, data scientists and subject-matter experts can now write their business logic for big data analytics using natural-language rules and decision tables. This makes the logic much easier to understand and the separation from the source language of the big data platform allows for more rapid change of the processing logic.
Nigel Crowther's articleThink big! Scale your business rules solutions up to the world of big dataapplies this pattern to Hadoop and ODM using services in the IBM hybrid cloud development platform, IBM Bluemix™.
For more information about the concepts that are introduced in this article, refer to the IBM Redbooks publication, Systems of Insight in the Digital Transformation Era, SG248293.
Likes before 03/04/2016 - 0
Views before 03/04/2016 - 3612