SAS IoT's contribution to analytic innovation

[zephstemle]

‘Turning a world of data to a world of intelligence’ is our mission in IOT Advanced Analytics R&D. We are enhancing our analytical functionality to go from streaming data to business value. In this video, about 17 minutes long, SAS' @GulEge highlights the latest and greatest functionality that we offer in SAS ESP and in SAS Viya to fulfill the promises of the connected world.

The presentation highlights new additions and innovations in these areas:

• Digital Signal Processing (DSP)
• K_T Charts
• Shapley Values
• Recommender System
• Kernel PCA (KPCA) Preimage
• Independent  Component Analysis (ICA)

Some new additions to DSP functionality include: Two Dimensional Fast Fourier Transform (2D FFT) and Peak Finder for signal transformation and analysis,  Short-Time Parametric Power Spectral Density Estimation (STPPSD) and Empirical Mode Decomposition (EMD) in the Time Frequency area. We demonstrate these additions in the domain of agriculture (beehive health and chicken coop) monitoring, gun shot sound identification and classification in the public health domain with acoustic data. We also have an example of Heart Rate Variability monitoring through health variables with Electrocardiogram (ECG) data.

K_T Charts are for monitoring both the volatility and drift of a process with multivariate high-frequency data, processing a moving window of observations at a time. Addressing a high priority use case in IOTP continuous health monitoring of mission-critical, highly reliable capital equipment for predictive maintenance. K_T charts functionality is based on Support Vector Data Description (SVDD) model. This offers early detection of system degradation. We demonstrate using turbofan engine data from airplanes.

The new Shapley Value plugin in ESP will help you answer these questions on streaming data in any domain: “Why is the model reporting 95% failure probability?”, “Which sensors are responsible for this result?”, and “Where should the factory manager send the technician?”

We also introduce Kernel PCA (KPCA) preimage with an example in facial image denoising and Independent Component Analysis (ICA), a blind source separation technique, in audio signal recovery.

Be sure to check the Analytic Innovation video for a more detailed introduction to this new functionality. Contact Gul Ege for questions and more information.

Watch the Video