Do (or don’t) believe the hype – part 1
You’re probably familiar with the Gartner Hype Cycle - a visual representation of different technologies plotted over time in relation to expectations of their impact and value. This year’s hype cycle for data science and machine learning came out in the summer and I thought it would be good to define some of the technologies - and their vendors - that are on the up before they reach peak hype.
Advanced anomaly detection
Anomaly detection is the identification of outliers which can be used to find things like malignant tumours in MRI scans or fraudulent transactions. The Civil Aviation Authority successfully applied advanced anomaly detection methods to its tail rotor data. Druva announced it is offering advanced anomaly detection capabilities with its “comprehensive set of trending charts and visualisations that track data activity.”
This is shorthand for automated machine learning. In the view of ml4aad, AutoML comprises off-the-shelf machine learning methods that can be used by people who are not experts in machine learning. According to two data professionals at AirBnB, AutoML is the automation of tasks such as exploratory data analysis, feature transformations, algorithm selection and hyper-parameter tuning, and model diagnostics. For them, AutoML was most useful for regression and classification problems that involved tabular datasets.
In the words of Logi Analytics, embedded analytics is the integration of analytical content and capabilities within business process applications. The analogy was made of business intelligence being a map to plan a journey, with embedded analytics being the in-car GPS navigation that guides the driver in real-time. Gartner describes it as the use of reporting and analytic capabilities in transactional business applications where those capabilities can run outside of the application. However, for Tableau, embedded analytics is about embedding data visualisations in an enterprise’s website or blog, which seems to be a sub-set of the aforementioned concept.
Citizen data science
Essentially, this is data science done by people for whom data science is not their day job. They might not be expert, but they can use tools to create data models and provide insights. A citizen data scientist could work in a related role, such as a business analyst, or they might be in a completely different field (potentially with no professional involvement, just a strong personal interest). And citizen data scientists work with data scientists in a way that is complementary, rather than competitive.
IoT edge computing and analytics
Edge computing is the ability for Internet of Things (IoT) devices to run complex computations and data processing on-site and integrate seamlessly with the cloud. According to IBM, IoT edge analytics means that devices have the capacity to pre-process data so that only anonymised, aggregated and privacy-compliant data is sent to the cloud for further analysis. IoT edge analytics can also reduce the time it takes to implement low-latency machine learning algorithms. Furthermore, applications that are designed to run on the edge are not affected by instances of low or irregular connectivity.
Part 2 will give definitions for the remaining six technologies that are getting hyped up.