By Cindy Gordon
Part one of this three-part examined how big data has changed our approach to crafting business questions and the expectations of what information we’ll have access to when we make decisions.
Today analytics is moving rapidly into business operating processes and practices. The term being coined to describe this new field is machine-to-machine process intelligence, which at its core is predictive analytics.
What fascinates us as data scientists is the analytics beyond the simple. Our focus on analytics is on the “predictive analytics edge,” where the analysis is based on real-time results and continues to learn from a baseline predictive model that is also flexible.
Example: Insurance Fraud Analysis
Consider insurance fraud analysis that was traditionally run, say, every two months. At that point the damage was done – the fraudulent insurance claim had already been paid. This approach was slow and passive.
Now insurance companies run in-database fraud analysis twice a day, catching fraudulent claims within hours, and increasingly within minutes. Whereas traditional systems were fine-tuned for transactions and batch processing, today we need to sense and respond to changing conditions immediately. Even correlating criminals living in certain neighbourhoods can add to risk insights for insurance providers.
Example: Cross-Selling Banking Services
In banking, every banking transaction you place becomes a permanent record of your customer buying profile. Data collected ranges from your credit purchasing trends, to investment profile, to banking wealth. The banking industry has been working diligently across its retail and wholesale banking divisions to get a more consolidated view of its customer profile attributes. The goal is to develop triggers on customer service approaches to provide:
- alternative services offers, whether in the transactional experience moments, or
- in upselling/cross selling potential solutions based on buyer profile and potential propensity for new solutions.
For example, predictive analytics on a customer’s credit card purchases identifies a stream of purchases over three months on baby products. Analysis shows there is no bank mortgage on file with their customer record.
A predictive trigger could recommend that a call from a mortgage advisor to make an offer and pre-empt the inevitable decision to seek a mortgage elsewhere.
In the past, we focused on structured relational databases, predefined relationships, and structured transactional data. Although those don’t go away, now we have volumes of unstructured data, untold relationships between the data, and new data coming in all the time. Hence we have more dynamic data, with multiple contexts, from many sources, both people and machines.
This shift in how we handle data reflects the fact that our world is complex and connected in ways we cannot imagine. The Internet has brought all of this complexity to us in a flood of data that is massive, diverse, unstructured for the most part, and rich in valuable information.
Further, what happened on the Internet is happening in corporations and governments. There are few rules and virtually no constraints. Every author of content can invent his or her own structure, create his or her own context, and tell his or her own story. Some of the data can be defined as well structured and some is semi-structured.
Summary
what we have learned is that in this vast data reservoir of diverse data sources, the most important driving force is making the right connections.
This means that predictive analytics approaches are critical to be applied. Methods from artificial intelligence, machine learning, advanced statistics and continual machine learning approaches are needed to enable these smarter connections.
One of the most powerful data connectors is the use of links or hash tags as this codification method can be used to mine and detect early signals or pattern inferences. We also have learning that linking allows us to connect diverse element data sources and that these combinations, together tell a story; further, the same story can be told in many different ways. In addition others can keep adding to the story patterns, so the predictive models continue to shift like different waves.
In the third and final blog post in this series, we’ll examine the rise of chief data scientists inside of businesses and other ways the data revolution has transformed company structure.
Information on the SalesChoice Predictions Insight Engine can be found here
“Brave ‘now’ world: Everything predictive and connected” was originally published on ITWorld Canada Business and is accessible here