The invention discloses a crowdsourcing software developer recommendation method. According to the method, first, a dataset is preprocessed, tasks with complete information are selected, and feature information (such as task development time, rewards, demand description, required skills, publishing time and a publishing platform) of the selected tasks are further extracted to construct feature vectors of the tasks; second, the tasks are clustered, and similar tasks are gathered together; third, for a new crowdsourcing task, features of the task are extracted in the same way, cluster tasks similar to the task are obtained, and further developers with high winning percentages serve as a preliminarily-recommended developer candidate set through a machine learning classification algorithm among the similar tasks; and last, the reliability of the developers is calculated in the developer candidate set, part of the developers with low reliability are removed, the incidence relation among thedevelopers is analyzed according to historical task participation conditions of the developers, and a developer competition network is established.