Abstract： The advantages of classical community detection algorithm based on label propagation include precision and time complexity. On the other hand, there are several disadvantages, such as oscillation, unstable result, tendency of big community. The nonrandom community detection algorithm based on label propagation (NCDAL) proposed by this paper improves the disadvantages of the classical algorithm by getting rid of its random procedures. There are three improvements in the NCDAL: when renewing the label, there is the special order; when summing up the labels, both current vertex and its neighbors are counted; and when selecting the label, the contribution function is defined to avoid random selection. It is approved by the experiments that the proposed algorithm not only has high precision, but also decreases random procedures of the classical algorithm.