An Oracle blog about Internet Of Things

Artificial Intelligence (AI) In Every Oracle IoT App

Guest Author

Viji Krishnamurthy Ph.D., Director, Product Management, Internet of Things Cloud

IoT Growth Driven by Industrial IoT  

Over the last couple of years, we have seen multiple reports on the growth of connected devices and the expectation of business value creation with IoT. While each study has its own way of measurement, we can be convinced of the following:

  • Number of connected devices are growing from ~20B now to over 75B by 2025.
  • While there is significant IoT market share in consumer applications, Industrial IoT is expected to be more than twice the value of consumer IoT with applications in Manufacturing, Transportation & Logistics, Agriculture and Healthcare.
  • Examples of Industrial IoT applications include Predictive Maintenance, Remote Patient Monitoring, Self-optimizing Production, Automated Inventory Management, Smart Meters, Connected Cars, Distributed Generation and Storage, and Fleet Management. Successful deployment of these applications require addressing complex intertwined business goals of multiple global business participants based on data sources that include devices and enterprise applications.

Typical industrial IoT deployment has thousands of sensors and enterprise data sources that span the supplier, manufacturer, logistics and warehouse participants. IoT’s benefit lies at the core of analyzing these data in real-time to derive business value. Due to the interdependency of data and participants’ actions along with the volume, velocity and variety of data, IoT’s benefit cannot be derived without AI. This is why, every Oracle IoT SaaS application is built with AI at its core.

“Data Scientist In A Box” - Painless AI

Oracle IoT SaaS applications are built for domain users such as factory managers, supply chain planners, fleet or warehouse managers. Given that the majority of Industrial IoT users aren’t from a data science background, AI in Oracle IoT SaaS applications are built with following specific design principles:

  1. Derive real-time decisions by automatically identifying influencing features from data sources. Automatically identify patterns in influencing features and their contribution to business event of interest.
  2. Adapt and improve decision science models by continuously analyzing device, data and user actions. Self-learn based on new data as well as adaptively learn based on domain user interaction.
  3. Automate decision making by triggering appropriate work flows in enterprise applications as well as edge devices to immediately derive benefit to business goals.

Typical IoT application in the market requires users to select the data sources, identify features and select the algorithm with algorithm specific parameters as well as the data preparation techniques. This places enormous expectation on the user from both skills as well as sheer volume of data thereby limiting the use of IoT. At Oracle, based on our interaction with industrial customers, we have designed our IoT applications with “Data Scientist in a Box” that automatically identifies influencing features and their patterns with respect to business events. When a domain user indicates the business event of interest for anomaly detection or prediction or action recommendation, Oracle IoT applications automatically evaluate appropriate AI pipeline of algorithms to generate a comprehensive data science model that is stored and scored with every new data. AI pipelines are designed with AutoML that selects appropriate chain of algorithms, Auto-Tuning that optimizes hyper parameters and Auto-Selection that helps to select most influencing features. As a result, AI in Oracle IoT applications helps the domain user to derive the value of IoT data directly and in real-time.

What Is Inside “Data Scientist in a Box”?

Industries have been applying data science for decades in many areas including supply chain planning, process optimization, fleet utilization and pricing. IoT data analytics differs sharply from these traditional data science efforts in the fact that it can’t be a periodic manual effort. Given the interdependency of devices and actions, IoT data is expected to change over time and therefore, IoT data analysis can’t be periodic and manual as in traditional data science efforts. Knowing this, AI in Oracle IoT applications is designed with both “Self-learning” and “Adaptive-learning” methods. AI in Oracle IoT applications track the accuracy of identified anomalies, predictions and recommendations over time to automatically trigger appropriate AI pipeline for re-training to regenerate data science models. In addition, AI in Oracle IoT applications adapt to the inputs provided by the domain users via Digital Twin. Specifically, domain users that include factory manager, equipment technician, maintenance manager and process engineer can record their input on IoT data as well as results of IoT data analytics via Oracle IoT applications’ Digital Twin interface. Oracle IoT applications have multiple user interfaces including AR, VR, mobile app and voice assistant devices. AI in Oracle IoT applications automatically adapt these user inputs to represent domain user knowledge in data science models thereby making the model closer to real behavior of industrial systems.

AI + Digital Thread = Automated Intelligent Workflows

IoT’s benefit in industries rely on real-time implementation of actions recommended by IoT analytics. If every IoT analytics recommendation requires user’s involvement then, we would have simply moved the problem from industrial systems to personnel thereby prohibiting us from deriving the benefit of IoT. Therefore, AI in Oracle IoT applications not only derives optimal actions, but also helps in automating work flows to execute those actions for agility. This is achieved in Oracle IoT applications in two specific ways to address the actions executed on the edge devices and enterprise software. If the action is executed on edge device, conducive to the capability of edge device, AI in Oracle IoT applications can off-load the real-time identification of events and action recommendation (called scoring) to the edge device through a combination of Digital Twin and Software-Enabled-Edge. This enables agile operational adjustment of edge system to maximize its performance while the Oracle IoT cloud application performs data science model updates in near-real-time to send to edge system as necessary. In case the action is executed on an enterprise software system, AI in Oracle IoT applications utilizes the data and knowledge of business entities in those applications via Digital Thread thereby creating global recommendations and triggering appropriate work flows. Specifically, AI in Oracle IoT applications utilizes data from Digital Thread to enterprise applications such as supply planning, maintenance, manufacturing, transportation logistics software and triggers appropriate exceptions, notifications, chat bot or service work orders as required to implement actions.


In summary, every Oracle IoT SaaS application implements AI to make IoT benefits immediate, comprehensive and practical. For more information on Oracle IoT applications and to request demo or to try for free, visit https://cloud.oracle.com/iot-apps.

Be the first to comment

Comments ( 0 )
Please enter your name.Please provide a valid email address.Please enter a comment.CAPTCHA challenge response provided was incorrect. Please try again.