This paper presents the class-centre vector based category algorithm, which is regarded as the basis of the text classification in case-based reasoning application. On the basis of the proposed algorithm, this method can be used to determine those cases'' affiliation, and as a result this can reduce the retrieval scope so as to reduce the calculation and improve the precision. We use the proposed algorithm in the traffic accident analysis and the corresponding early warning system. The experimental results and the analysis prove the feasibility of the approach, and the existing problems and further works are also discussed in the end.