An ensemble method: Using CBR and the Inverse Problem to Identify Deviations in Business Process

Abstract

This paper presents a work in progress research on how the Inverse problem (IP) can be applied to enhance Case-based Reasoning (CBR). It develops a novel model that extracts the capabili-ties of CBR in solving problems in domains where there is scarce or no mathematical founda-tions and combines it with inverse problem techniques (IPTs). IPTs have registered numerous successes in science and engineering fields since many important real-world problems can be solved via an application of IP. The areas of the problem space that are not represented in a case base will be created using Machine Learning (ML) algorithms whereby problems that are not captured by the current case base are generated and matched with a known case to ascertain the case genuineness. This is an inverse problem because it takes the output generated from the solutions of CBR probabilistically and predicts relevant cases for the case base. It will be ap-plied to a drilling process. We will present preliminary results which formulate parts of the forward problem. A simulated sample data obtained from relevant literature was used to test the model performance against some selected metrics. This approach will be useful not only for increasing the problem coverage of the case base but also in creating cases with rare solutions, improve the performance of an existing Case-based reasoning system by automatically generat-ing synthetic cases hence boosting the CBR.