The code reviewer assignment problem affects the reviewing time of a source code change. To effectively perform the code review process of a software project, the code reviewer assignment problem must be dealt with. Reviewer recommendation can reduce the time required for finding appropriate reviewers for a given source code change. In this paper, we propose a reviewer recommendation approach based on latent Dirichlet allocation (LDA). The proposed reviewer recommendation approach consists of a review expertise generation phase and a reviewer recommendation phase. The review expertise generation phase generates the review expertise of developers for topics of source code changes from the review history of a software project. The reviewer recommendation phase computes the review scores of the developers according to the topic distribution of a given source code change and the review expertise of the developers. In an empirical evaluation of five open source projects, we confirm that the proposed reviewer recommendation approach obtains better average top-10 accuracy than existing reviewer recommendation approaches.