The Record - Issue 18: Autumn 2020

112 www. t e c h n o l o g y r e c o r d . c om I n my previous articles for The Record , we talked about the first two steps in the pro- cess of gearing up on how you exploit data – collecting the data and making it coher- ent, then starting to unlock its potential with visualisation, applications, analytics and artifi- cial intelligence (AI). In this article, we’ll drill down into the higher-order techniques for extracting knowledge and making data action- able, along with other aspects of increasing the value of your data. First, I wanted to highlight a couple of key ana- lytics platforms fromMicrosoft itself, Stream and Synapse Analytics. Stream Analytics is designed to provide real-time analytics for inbound data streams, such as those from connected vehi- cles, the plant floor, or in a mobility or freight- as-a-service system. Being serverless, Stream Analytics is easy to deploy and configure and integrates with Azure internet of things (IoT) technologies. Standard Query Language (SQL) syntax can be used for ease of query generation, or you could insert your own code, or even take advantage of integrated machine learning (ML). It can be run at point of ingest, higher up the stack on the results of a data subscription, or nearer to data sources with Azure IoT Edge. This flexibility allows one solution to be used for a wide variety of use cases and needs. While Synapse Analytics also can support real time data and streams, think of this as your tool for managing a broad set of needs for historic data analysis. Synapse accelerates time to insight by simplifying and integrating the acquisition, manipulation, management and extraction of data, leveraging well-known para- digms and tools such as SQL and Apache Spark. Synapse Studio provides a one-stop-shop for everything the data engineer needs and can be embedded into continuous integration/contin- uous delivery approaches. The goal of collecting a large and represent- ative set of coherent data is to automate the unlocking of meaning and insight. The high- est yield here can be obtained with cognitive approaches leveraging AI and ML. Scepticism is sometimes expressed about the viability of these solutions, but it is worth noting that even in traditionally challenging areas Microsoft AI solutions have achieved and exceeded parity with human performance in most areas of per- ception – see the ‘Microsoft AI breakthroughs’ illustration for examples. With AI ready for prime time, there are hun- dreds of ways in which it can be applied across the automotive industry. But our focus here is on gaining knowledge from large data sets. In this space, the most useful scenarios are those where machine learning is used to spot pat- terns and derive predictions from datasets. For example, based on a sufficient set of telemetry and known outcomes, a predictive maintenance model can be derived, validated, and put into production to minimise downtime and waste. Information can also be inferred from previ- ously unexploited but extractable correlations to other real-world properties. A good example The power of data AI, machine learning and a coherent data platform can help unlock the value of your data J OHN S T ENL AK E : M I C ROSOF T V I EWPO I NT “Microsoft AI solutions have achieved and exceeded parity with human performance in most areas of perception”