Traditional attribute reduction algorithms cannot dynamically update the results of the reduction effectively. The existing incremental attribute reduction algorithm compensates, to a certain extent, the defects of the traditional attribute algorithm, but most of them focused on either categorical attributes data or numerical attributes data simply. In order to process the dynamic data with hybrid attributes, and by combining the advantages of the neighborhood rough sets model that it processes hybrid attributes effectively, a knowledge granularity calculating method based on the neighborhood rough set model is proposed. Through dynamically updating the neighborhood particles to incrementally get the knowledge granularity. According to the knowledge granularity, an incremental reduction algorithm is further developed. The validity of the proposed algorithm is demonstrated by experimental analysis.