Following the release of the Essential Forecasting app in January, Siemens is now extending the capabilities of their AI for Everyone product line with Essential Anomaly Detection functionalities. These releases are part of a larger plan of Siemens Digital Industries to develop a range of no-code, out-of-the-box applications for easy, efficient and transparent predictive analytics on IoT data for which they partnered with the time-series machine learning experts of Tangent Works.
AI for Everyone is accessible from the user interface of Operations Insight within the MindSphere IoT platform. It offers an end-to-end, out-of-the-box model building strategy that identifies relevant features and generates a unique GAM (Generalized Additive Model) for each time series forecasting or anomaly detections use case. To accomplish this, AI for Everyone leverages TIM from Tangent Works that harnesses the mathematics of Information Geometry to expand the original explanatory variables into many different features and then very efficiently reduces them into a unique GAM model.
Essential Anomaly Detection capabilities comprise:
- Define “Use Case” to manage both Forecast and Anomaly Detection experiment jobs and linked dataset MindSphere based on asset model.
- Build Anomaly Detection model via parameters configuration of training dataset, detection KPI and predictors, Perspective & Sensitivity, Normal behavior, Data imputation and so on
- Detect anomaly spots from in-sample dataset based on built model
- Visualize model building and detection results of anomaly spots, perspectives, predictor importance, historic model building and detection records, etc.
These first two functionalities of the AI for Everyone product line powered by TIM are just the start. The new capabilities will be presented at the upcoming Hannover Messe and several more advanced and use case specific apps are already being developed and scheduled to be released later this year.
Visit the Siemens Mindsphere releases notes for more information.