Adding Probability Based Fusion Criteria for Performance Evaluation in Multiple Information Systems
Hong-Dar Lin · 朝陽學報 · 1997
Most current information fusion techniques focus on adding as many different information sources as possible, and then observe the performance of the final fused result. They generally do not concern whether a new information source is useful or not until it is added into the fusion system. The statistical meta-analysis offers a set of quantitative techniques that permit synthesizing a variety of independent information sources. Using the statistical meta-analysis, this paper presents a method that can provide fusion criteria to foretell the effect of adding a new information source in terms of statistical type Ⅰ,type Ⅱerrors and an expected error cost before the source is actually combined. Implementation results confirm the highly accurate prediction power of the proposed models.