Unequal Binning Distorts Conflict Measurement in Evidence Fusion

Written by

in

A study found that when unequal-width bins are used for continuous variables, the Dempster conflict coefficient K incorporates geometric bias caused by bin positions and widths, affecting judgments about disagreement among different scoring sources. The study analyzed more than 13,000 double-assessment records and used 37,975 records from 2024 to compare four BPA construction methods and six combination rules. No fusion method was significantly better than the best single source. The researchers proposed calculating baseline conflict from two identical readings at the same position and correcting for it accordingly. They also suggested scanning the entire bin range to check whether conflict is flat.

Source:BIOENGINEER.ORG · Read Original

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *