Efficient Privacy Aware MCDM

Multiple-criteria decision-making (MCDM) is a well known technique that allows to structure and solve decision and planning problems involving multiple criteria. Typically, such problems do not have unique optimal solutions. Therefore, it is essential to use decision-maker’s preferences to differentiate between solutions. In order to find the decisionmaker’s preferred solution(s), MCDM techniques need to compare potential solutions against decision-maker’s preference (e.g., ideal solution, preferred priorities / weights). However, when the number of solutions get increased, the number of comparisons that need to be performed also get increased.

In this project, we propose to use machine learning techniques to build predictive filters in order to eliminate solutions with least probability to get selected as preferred solutions. As a result, the number of comparison need to be performed get reduced. We compare the accuracy and performance of our approach with two other approaches: 1) a heuristic approach and 2) the brute force (i.e., traditional approach that solves MCDM problems). Our proposed approach is efficient in three ways. Firstly, it drastically reduces the potential solution comparisons (i.e., computationally efficient). Secondly, it reduces the decision making latency (i.e., decisions are made much faster). Thirdly, it reduces the network traffic in situations where the decision space is very large and stored in a centralised database serving distributed compute nodes. However, above efficiencies are achieved by sacrificing the overall accuracy up to some level. Therefore, we recommend our approach as a tradeoff mechanism which would be suitable in curtain contexts where achieving above efficiencies are more important than accuracy.

Team

Funding

Outcomes