Predictive Intelligence technology The pioneering methodology known as Predictive Intelligence is a combination of proven mathematical techniques that, when applied to a specific system of interest, allow for monitoring the current conditions as well as viewing predicted changes in key parameters. Further, this approach to modeling dynamically updates its forecasts by comparing current conditions to previous predictions, effectively providing an estimate of quality to the predictions.

A main feature of Predictive Intelligence is its capability for data fusion; assimilating real-time data into the overall system model. By incorporating system observations as they occur, the user receives a two-fold benefit: a real-time view of their system of interest and a predictive capability which continually improves upon predictions. In the past, using observable data in model representations was difficult and provided little insight into the underlying mechanisms. Either the data did not fit the model variable criteria or the computational power limited the expanse of the model. At the same time, a model without observations provides no verifiable conclusions. This resulted in a modeling/simulation environment in which analysis of each simulation was required in order to tweak the model for the next iteration. This methodology works fine in a design environment but in situations where the system of interest is rapidly changing results were too delayed to be of use. The objective is to enable inclusion of all observable data, in real time whether the data is complete or not, and to prevent the model's output from diverging from reality. Based on an advanced technique known as data assimilation and coupled with a technique known as Kalman filtering, these objectives are attainable. However, it requires expertise in physics, mathematics and modeling in order to develop such a Predictive Intelligence system. With the advances in computational power, more complex models with millions of variables are now possible.

At the core of our predictive intelligence methodology lies the system model, which can be either first-principles and/or empirical depending on the environment modeled. The primary mathematical techniques reside in the data fusion element, which delivers data to the model for computation, and solves the observation density issue, when there are too few observations to run the model resulting in misdirection and incorrect results. The overlying element which couples the physics models and empirical models, for the most complete representation possible, and ensures that the output will not diverge from reality is the confidence level determinate component. Not only does this constituent fill in the blanks for missing observations in order to provide a real time view of system, it also provides a defined confidence for each predicted parameter. These primary components are supplemented with a variety of state-of-the-art techniques, either for algorithmic development or for incoming data characterization.

Our technological and developmental framework is known as PredictIT.