Jesteś tutaj

Models and Methods of data conflict resolution in belief functions theory

Nowadays, many mathematical theories are available to manage data imperfection in data fusion process. The belief functions theory is acknowledged to be one of the most used theory in this field because of its ability to represent and handle various aspects of imperfection. However, it is claimed that this theory output counter-intuitive results when fusing highly conflicting information using Dempster’s combination rule. In this regard, we will show in this seminar methods to solve this problem by incorporating source reliability degree to the fusion process. We also propose a new model for source reliability degree and propose new methods that allow us to estimate its values under different challenging scenarios (luck of data training, large volume of datasets and streaming datasets).
Prelegent (ISI): 
Rodzaj seminarium: 
czwartek, Grudzień 15, 2016 - 14:00