KPCARST: a method of weakening conflict evidence based on kernel principal component analysis and relatively similar transformation

How to deal with conflict evidence is an essential issue in Dempster-Shafer evidence theory (DST). In the Dempster combination rule, a high anomaly can produce counterintuitive phenomena, so scholars have proposed many effective methods to weaken the anomaly. This paper proposes a novel anomaly evidence detection framework based on kernel principal components analysis and relative similarity transformation (KPCARST) to address this issue further. The new method can weaken anomalies’ influence on the evidence fusion results and achieve more effective results than existing methods. A numerical example verifies the effectiveness of the new method.