You are here
Recognition of Errors and Validation of Input for Self-healing Entry (REVISE)
Phone: (215) 542-1400
Phone: (215) 542-1400
The objective of REVISE (Recognition of Errors and Validation of Input for Self-healing Entry) is to augment legacy information systems with data healing functionality, initially in support of the Naval Depot Modernization and Sustainment Program and ultimately in other military and non-military domains. The Phase I effort successfully demonstrated the feasibility of developing a fully functional REVISE. Specifically, based on a representative maintenance dataset, multiple ML (Machine-Learned) models were constructed to identify potentially incorrect data elements and recommend corrections. Various classes of algorithms were evaluated to determine the relative reliability of these ML-based context models in the prediction of Naval Aviation Logistics Command/Management Information Systems (NALCOMIS) errors. Analysis of the results from this feasibility prototyping effort showed that ensemble Classification and Regression Tree (CART) algorithms provided the highest performance and accuracy (i.e., mean accuracy over 92%) in the healing/correction of errors for various fields within the example dataset. As part of this prototyping and analysis process, a big data ML context model construction pipeline was developed. In parallel, user interface and enterprise-level functional concepts for the REVISE system were developed, aimed at providing a usable access framework to support a variety of potential user personas within the larger information system context. Based on the successful Phase I demonstration of the REVISE data healing approach and the development of associated functionality concepts, CHI Systems proposes to extend the Phase I software and develop a fully realized REVISE product. Under Phase II, a Minimal Viable Product (MVP) of the REVISE system for data healing will be developed, including: user interface and visualization capabilities, the infrastructure to support integration with existing systems, and an automated pipeline for the development of data healing context models for domains beyond the Phase I NALCOMIS maintenance dataset. The product developed in Phase II will be tested with target user populations to establish MVP efficacy and usability, and support continuing refinement of software releases. Phase II will complete the productization of REVISE and demonstrate functionality in a Naval maintenance environment.
* Information listed above is at the time of submission. *