A prestudy of the accuracy of three predictions in an ICU setting.

Purpose and goal

Our purpose and goal was to analyze technical opportunities and business opportunities for a greater project, see which parameters we can accomplish the best predictions on, assess a potential collaboration with Daintel and identify other market players. The conclusion is that there are great technological and business opportunities. We have not been able to evaluate the predictions on Daintel data due to invalid data formats and as of now we cannot work with them. Other players are identified.

Results and expected effects

Our result has been very favorable, as we have reached a technological breakthrough. We expect to deploy complex predictive models much faster than accustomed. Considering the expectations on the prestudy the project has taken another path due to invalid data formats of the ICU data. Therefore the path chosen has diverged in terms of result and time. The net end result will be the same because the delay will be compensated for by a greater achievement in the technological development.

Approach and implementation

We developed a number of generally applicable mathematical tools to manage time series data. Other data sets were used. When data from Region H arrived a full error detection work began to understand the complexity and unlock data. Daintel can solve the problem first in 2016 and Region H expects change of vendor. The developed tools have been verified and validated, albeit not on high dimensional ICU time series data. Appropriate time series data is due in weeks to accomplish the prestudy.