One popular metric for gauging the robustness of a community’s structure in community discovery is the modularity function. Using a genetic algorithm and a random walk model, we present a new way to encode network community structures and an approach to community detection that maximizes modularity. To begin, the initial adjacency matrix was weighted using the results of applying the random walk model to the calculation of node similarity. A weighted network link topology was subsequently encoded into a chromosome in accordance with the closest neighbor-based similarity representation approach. By adjusting the structure of a set number of chromosomes and then computing their modularity, the genetic algorithm was able to find the optimal modularity value, which is related to the number of communities and their structure. On several actual social networks, we put our technique to the test. The innovative approach achieved higher modularity value in comparison to multiple state-of-the-art methods. The results obtained using the suggested strategy are more realistic because it does not necessitate a predetermined number of communities when community partitioning begins. In this case, the ideal community layout and number of communities are calculated automatically.