With the proliferation of complex networks, a large number of effective community detection methods have been developed. Of these, the label propagation-based approach for community detection has proven to be very suitable for community detection. However, it should be noted that this approach is stochastic and unstable. To address these issues, a CILPA method is proposed based on label Propagation and node information to identify a more rational community. First, we obtain the local influence reflecting the local spatial location of nodes based on their neighbourhood relationships. The importance of the nodes is subsequently obtained by combining topological information and local influence. Secondly, a label propagation strategy with preference selection is introduced to obtain stable overlapping communities based on neighbourhood information. More importantly, a label checking strategy is articulated to reduce the impact of sparse connections. Extensive experiments on the LFR benchmark network and real-world networks demonstrate that the approach is both stable and effective.