A Data Classifier Based on Maximum Likelihood Evidential Reasoning Rule

In Dempster-Shafer evidence theory (DST), some classical evidence combination rules can be used to fuse the multiple pieces of evidence, respectively abstracted from different attributes (features) so as to increase the accuracy of multiattribute classification decision making. However, most of them have not yet considered the interdependence among multiple pieces of evidence. The newly proposed maximum likelihood evidential reasoning (MAKER) rule measures such ubiquitous interdependence by introducing correlation factors into evidence combination. Hence, this paper designs a MAKER-based classifier to mine more correlation information for data classification. Finally, some numerical analysis (classification) experiments are carried out using five popular benchmark databases from the University of California, Irvine (UCI) to illustrate that the refined measure for evidence interdependence can aggregate the fused probability (belief degree) into real class label of a sample and further improve classification accuracy.