Prescriptive Analytics rely on the predictive models for optimized exploitation of current data. The goal of Prescriptive Analytics is to specify a preferred course of action, recommend a decision, provide decision support information, or provide an automated command.
For the Prescriptive Analytics implemented in IT systems, the decisions usually require human assessment before a mission-critical action is taken. Sometimes, the decisions may be automatically implemented as automated control and optimization logic. The mission-critical on-line decision logic would be normally implemented in OT systems.
Mitek provides Predictive Analytics, such as Digital Twins, as stable Analytics Engine products. At the same time, Prescriptive Analytics functions usually have to be adapted to realize desirable outcomes and for specific business cases. The Prescriptive Analytics rely on predictive models to optimize the impacts of possible decisions and provide value-added recommendations.
Mitek Analytics’ experience with prescriptive analytics includes optimization of power load forecasting for electrical utilities, energy efficiency performance trending for commercial aircraft, prognostics and health management (PHM) for jet engines, asset management analytics for power systems, anomaly monitoring analytics for aircraft and spacecraft, and power generating turbines, reliability analysis for military equipment fleets, and other decision support applications. Mitek has developed and demonstrated Prescriptive Analytics for several common business cases such as:
- Condition Based Maintenance (CBM) to reduce service cost
- Forecasting and predictive trending for monitoring and decision support
- Asset health monitoring to reduce downtime and improve customer loyalty
- Analysis and optimization of sustainment and logistics
- Optimization of energy and fuel efficiency
- Stochastic optimization of electricity trading
- Tracking product usage, e.g., aircraft structure usage analytics
- Risk estimation and management
- Reliability analytics, identifying supplier or operational issues
Our Prescriptive Analytics use several advanced analytical functions developed in earlier engagements. The experience includes the functions for energy efficiency analytics, analysis and reporting of fuel efficiency for the fleet, performance trending, multivariate anomaly monitoring, fault isolation, Condition Based Maintenance (CBM) analytics, risk analytics, and reliability analysis. The multi-level predictive models can be used as a basis of many other analytical applications.
Mitek has developed a library of algorithmic methods that support the above functions including:
- Multilevel multivariate statistical process control logic for anomaly detection
- Multilevel (longitudinal and cross-sectional) trending for asset fleets
- Fault isolation for parametric and event data to evaluate root causes of anomalies
- Multilevel analysis of discrete events for reliability
- Multilevel analysis of extreme events for risk assessment
- Stochastic forecasting tools based on multiple quantile regression