Abstract: The evaluation of algorithm scheme in radar tracking may be achieved by multi-attributes fusion via DS theory. However, the missing of some index in practice will result in high conflicts among evidences and make the results of classical DS-based fusion methods paradox. By analyzing three kinds of incomplete tracking information, this paper proposes a new DS-based fusion method, in which the evaluation of algorithm scheme is treated as the group evidence and the fuzzy distance between a single evidence and group evidence is adopted as the weight in DS combination. Experiment results show that the improved DS-based fusion could resolve the problem of evidence portfolios in radar tracking.