Five tips for understanding Internet of Things Analytics

Internet of Things (IoT) analytics refers to the analysis of data in the scope of IoT applications, notably through the processing of data streams from numerous sensors and Internet-connected devices. IoT analytics are an integral element of all non-trivial IoT applications in both the consumer (e.g., healthcare, smart home) and the industrial (e.g., energy, manufacturing) IoT spaces.

Their typical use in the scope of IoT applications is two-fold:

  • Knowledge extraction and presentation: They are used to generate knowledge from IoT data, which is accordingly visualised in appropriate reports and dashboards.
  • Driving decisions and actuating service: They analyse data in order to drive decisions and actuating services in an optimal way.

In the following paragraphs we highlight five important facts about IoT analytics, which facilitate the understanding of the main technical elements, but also of their business value.

 

  • IoT Analytics are essentially BigData Analytics

IoT data feature the four Vs (Volume, Velocity, Variety, Veracity) of BigData[1], as they are typically characterized by:

  • High-data volumes, since in several cases they usually collect and process streaming information from thousands of sensors.
  • High-velocity streams, since they usually involve streaming data that are collected and in several cases processed in real-time.
  • High-Variety, since it is usual to interface and leverage data from heterogeneous sensors and internet-connected devices.
  • High-Veracity, as sensor data are typically noisy and prone to errors and the unreliability of the devices.

As a result, BigData techniques and tools are applicable to IoT Analytics, including tools and models for data mining and machine learning, as well as streaming engines and databases. Nevertheless, adaptations of these tools to the nature and special characteristics of IoT data (such as their dynamic and streaming nature) are needed.

 

  • IoT Data is a perfect playground for applying deep learning (“Deep IoT Analytics”)

The nature of IoT data (i.e. multimedia, multimodal and noisy data) makes deep learning methods ideal for IoT analytics. Deep learning methods employ neural networks and fall in the realm of AI (Artificial Intelligence). Their application in IoT analytics is conveniently called “Deep IoT Analytics” and involves transforming the raw data to insights and actionable knowledge, as well as creating effective representations for machines and human users in order to deliver enhanced automation. Note that deep learning methods are favored over conventional machine learning, when it comes to extracting and understanding previous unknown patterns from complex data sources, as is the case in complex IoT problems. Deep Learning is one of the most trending BigData topics for 2017[2] and IoT provides a privileged domain for their application.

 

  • IoT data are usually streaming data

The majority of IoT applications process streaming data rather than batch datasets. As IoT data feature high ingestion rates, their processing is in most cases supported by streaming analytics frameworks. Such frameworks are provided as part of the popular open source framework for handling data streams, such as Apache Spark (spark.apache.org), Apache Flink (flink.apache.org) and Apache Storm (storm.apache.org). These frameworks provide capabilities for fast in-memory processing of data streams, while at the same time integrating libraries for data mining and machine learning.  While IoT data are essentially Big Data, their streaming natures render the above streaming frameworks more appropriate for IoT processing when compared to the conventional Hadoop[3]/MapReduce combination, which is typically used for batch processing of large datasets.

 

  • IoT Analytics can be deployed both at the Cloud and at the Edge of the IoT network

IoT applications have varying needs for analyzing data, including different performance, latency, availability and deployment requirements. In principle, IoT analytics systems are Cloud-based, since they are deployed within Cloud computing infrastructures in order to leverage the scalability, capacity and quality of services of the Cloud. Nevertheless, several applications cannot afford the latency associated with processing data in the Cloud, while being in need of filtering data close to the data sources in order to reduce the amount of information that is transferred and stored in the Cloud. Such applications are increasingly supported by the emerging edge/for computing paradigm, which makes provisioning for processing IoT data close to the field. Overall, IoT analytics can be deployed at the Cloud and at the edge of the network in various configurations as needed in order to meet the latency, performance, scalability and quality of service requirements of the deployment. Furthermore, analytics functions at the Cloud and at the edge are likely to be interacting (i.e. Cloud2Edge and Edge2Cloud interactions) as part of large scale IoT deployments. These interactions enable events derived at the edge to trigger cloud processing and vice versa. Note also that in addition to the presented streaming analytics frameworks for the cloud, frameworks for data analytics at the edge are emerging, such as the Apache Edgent framework[4]

 

  • IoT Analytics are key to realizing IoT’s business value

The efficient processing of large volumes of IoT data will be one of the major components of IoT’s business value in the coming years. According to a recent study by McKinsey & Company[5], most IoT data are not used currently: As an example only 1 per cent of data from an oil rig with 30,000 sensors is currently examined. Data used today are mostly exploited for anomaly detection and control, rather than for optimization and prediction, which provide the greatest value. Some of the emerging “killer” applications of the industrial IoT (e.g., quality control and predictive maintenance) are enabled by advanced IoT analytics.

IoT analytics will be an integral part of future IoT applications, as well as one of the main enablers of IoT innovation. In the coming years we expect to see an effective collaboration between IoT experts and data scientists in order to realize the innovation and business value potential of IoT analytics.

 

[1] http://www.ibmbigdatahub.com/infographic/four-vs-big-data

[2] http://www.kdnuggets.com/2016/12/ibm-predictions-deep-learning-2017.html

[3] hadoop.apache.org

[4] edgent.apache.org

[5] http://www.mckinsey.com/business-functions/digital-mckinsey/our-insights/the-internet-of-things-the-value-of-digitizing-the-physical-world

 

All information/views/opinions expressed in this article are that of the author. This Website may or may not agree with the same.

Click here to opt out of Google Analytics