This paper considers a paternity and kinship analysis system(PKAS) being currently used in real sites. A knowledge-based expert system is proposed to improve the performance of PKAS in terms of accuracy, speed, training time, and satisfaction, which are common measures for evaluation. The knowledge base, one of the most important components in the knowledge-based expert system(KBES), consists of a rule made from random matching algorithm, decision rules of allele types and guide rules of options. The last two rules are learning incrementally from sample data. The results show that PKAS armed with the expert system ensures the better performance with regard to these criteria than the existing system. Especially as far as speed is concerned, as the sample size increases, it outperforms the existing one. As the number of samples increases, while processing time increases nearly exponentially in the existing PKAS, it does linearly in our proposed system.