BT

Facilitating the Spread of Knowledge and Innovation in Professional Software Development

Write for InfoQ

Topics

Choose your language

InfoQ Homepage News Data Science at the Intersection of Emerging Technologies

Data Science at the Intersection of Emerging Technologies

Bookmarks

Kirk Borne, principal data scientist at Booz Allen Hamilton, gave a keynote presentation at this year’s  Oracle Code One Conference on how the connection between emerging technologies, data, and machine learning are transforming data into value.  

Borne explained that with IoT, the value is not in all of the data produced, but instead in the contextual knowledge that sensors give us about the world. With IOT and context knowledge, the goal is to discover and deliver value from data. Value from data is achieved by going from understanding to actions. Emerging technological innovations like AI, robotics, computer vision and more, are enabled by data and create value from data. Borne says that AI is not artificial intelligence but "actionable intelligence", and data is the fuel for insights for actionable intelligence.

In his talk, Borne went into four broad types of discovery from data: class discovery, correlation and causality discovery, anomaly discovery, and association discovery.

Class discovery learns the groupings and boundaries that separate groups in data. An example of class discovery is disease diagnostics given in lab measurements. Learning not only the groupings but also what distinguishes them is the real power with class discovery.

Correlation discovery finds dependencies in data. This is predictive discovery: if x correlates with y, then given x predict y. As you add data, you get insights as to the causal information about why something is happening, and this becomes causation or prescriptive discovery, which is given y find x. Discovering causal variables or prescriptive analytics allows you to predict and change outcomes. Examples of getting value through this type of discovery are fraud and machine monitoring.

Anomaly discovery finds regions that the data avoids. The a-ha moment answers the question: what is the data telling me about why it should not be in this spot?

Association discovery finds interesting associations, links in a graph that are indirectly connected. For example, a communicates to c through b. Examples of getting value through this type of discovery are recommendations, marketing attribution, and illicit money transfers.

Borne also discussed analytics, the products of data science, and the five levels of analytics maturity: descriptive, diagnostic, predictive, prescriptive and cognitive. Descriptive analytics reports what happened, which sounds boring but is essential for applications like auditing. Diagnostic analyzes why this is happening, why this changed under this condition. Predictive analyzes what will happen. Prescriptive analyzes how to optimize what happens. Cognitive, the highest level, analyzes what is the right action for this data in this context. Analytics by design focuse on products that generate value from data. An example is a chatbot recommendation engine which answers customer FAQs. This can add value by giving a better customer and employee experience.

Lastly, Borne went on to explain dynamic-data-driven applications which combine sensor measurements, machine learning inference prediction, and action. Borne said there is a combinatorial explosion of possible connections among emerging technologies like Robotics, computer vision, IoT, etc, which are building dynamic data driven applications not imagined before, such as:

  • Robotics combines AI with actions to sensory data. Examples include prosthetics warehousing, and manufacturing.
  • Augmented reality combines AI, computer vision and image recognition with actions to superimpose a computer generated image on the user’s view of the real world. Examples include medical procedures and trying out clothes without putting them on.
  • IoT sensor data combined with AI examples include: health devices, manufacturing, connected products and farming.
  • Machine Learning is also being combined with data to create smart data. Examples include metadata, taxonomies, and breadcrumbs.

Borne concluded his talk with a Larry Ellison quote:

Our mission is to help people see data in new ways, discover insights, unlock endless possibilities.

Rate this Article

Adoption
Style

BT