When it comes to IoT, will hybrid analytics prevail? According to Dave Shuman, IoT & Manufacturing Leader at Cloudera, in 2018 two major shifts will happen with regard to IoT. Shuman predicts we will continue to see IoT being less about connectivity and more about data analytics and machine learning. He also predicts 2018 will be the year that there will be an increasing focus on intelligence at the edge, and the best approach for many companies will be a hybrid analytical approach, which brings out the best of both the data hub and the edge architectures.

Prediction #1: IoT will be less about connectivity and more about data analytics & machine learning

The focus for IoT will shift from connectivity and onboarding to more on analytics and machine learning:

Based on all of the data that is continuing to be generated by connected devices, advanced analytics and machine learning will play more crucial roles in IoT, as organizations will increasingly focus on driving automation and intelligence into their operations. We will see an increasing adoption of machine learning at scale to enable concepts such as pattern recognition, anomaly detection, and predictive modeling using petabytes of sensor data that are generated from IoT use cases. Organizations will increasingly employ data scientists to build and deploy machine learning models with the goal to make the ecosystem more intelligent in order to predict and respond to situations including — predicting failures or downtime based on a equipment signatures, suggesting specific maintenance requirements, or making changes to the operational parameters based on asset performance.

Prediction #2: Increasing focus on intelligence at the edge, but ‘hybrid analytics’ will prevail

Even though there are some definite advantages in processing some of the data at the edge, organizations still will need to bring a lot of their IoT data sets into a centralized data store to drive advanced analytics and machine learning. Organizations are bringing in IoT data sets and adding context to sensor data by combining it with other enterprise data sources, such as CRM, ERP, supply chain systems, or data from external sources, such as weather or traffic data, to drive compelling insights.

The best approach for many companies will be a hybrid analytical approach, which brings out the best of both the hub and the edge architectures.

For example, a manufacturer might do some local filtering and processing at the edge, but still send data back to the hub for context enrichment, predictive analytics, and building machine learning models. Finally, machine learning models and lessons learned at the central hub might be pushed back out to the edge, so that intelligent decisions and additional adjustments to manufacturing processes can be made, in real time, closer to the source.

Dave Shuman is a subject-matter expert at Cloudera and has an extensive background in IoT, business intelligence applications, database architecture, logical and physical database design, and data warehousing.