Analytics and Data Summit 2020

There are increasing requests from customers for advanced prognostic ML techniques for predictive maintenance use cases for many industrial segments Oracl serves. The greatest challenges for applying any types of Oracle (or 3rd party) ML techniques for time-series prognostics do not have to do with the technique selected (including NNs, SVMs, Kernel Regression, MSET2, etc). Rather, the greatest challenges have to do with sensor and signal issues that occur across all IoT industries... e.g. disparate sampling rates, missing values, quantized signals, decalibration bias in sensors, and many others that will be presented. OracleLabs has developed a suite of Oracle-patented algorithms that detect and autonomously correct these and other common sensor and signal challenges. These Intelligent Data Preprocessing algorithms have been incorporated into a new Data Preprocessing Pipeline (DPP) tool/framework that will be presented with numerous examples from real industrial use cases.