Micromodels

The application of Artificial Intelligence to deliver predictive intelligence in real-world, dynamic operational environments is complex. There are many challenges and considerations that require sophisticated technology and techniques to overcome.

The real world is an evolving state. There are various factors that must be considered to deliver accurate and effective predictive intelligence. Some of these factors have a correlation that can be modelled and explained mathematically. In these cases, statistical and stochastic models work. However, there are also occurrences of what we call ‘black swan’ events. These are events that do not have a mathematical correlation, they are random. However, when they occur, they have a material impact on future state and therefore, must be considered in effective predictive models.

Entopy’s micromodel technology focuses on breaking large, predictive problems into smaller, specific ‘chunks’, targeting Artificial Intelligence models to specific parts of the problem and networking the outputs with real-time, event-base data. This creates a network of computational, statistical, and stochastic models to interoperate, creating a dynamic network capable of delivering predictive intelligence.

The resulting network is effectively a multi-layer-perceptron (MLP) but with each layer comprising multiple independent models. Entopy’s has developed a library of general micromodels and technology that enables networks of many models to be deployed in days.

There are many benefits to Entopy’s micromodel technology. First, its ability to deliver dynamic predictive intelligence which is key for delivering such capabilities effectively, in real-world, dynamic environments. Second, the network of independent models supports data segmentation enabling effective deployment in multi-stakeholder ecosystems. The data is used within individual models, with the derived outputs networked together ensuring granular segmentation and effective isolation of raw data. Critically, the predictive outputs can be understood and interpreted, enabling users and operators to explain the predictive outputs and recommendations. This critical capability helps to build confidence with operators and influence decision-making and action.