The invention provides a dynamic high-risk customer-group detection algorithm. The algorithm is used for detecting risk degrees of new users belonging to different customer groups, and includes the following steps: i, using a clustering algorithm to build the groups for multiple historical users on the basis of multiple pieces of historical data of the multiple historical users to form the multiple customer groups, wherein the clustering algorithm is related to types of the historical data; a, determining risk degrees of all the customer groups on the basis of an evaluation index, and determining that the customer groups of which the risk degrees are greater than a risk threshold value are high-risk customer groups; and b, determining similarity degrees between the new users and the high-risk customer groups on the basis of a distance function, and determining that the new users of which the similarity degrees are greater than a first similarity degree threshold value are risk users, wherein the distance function is related to types of user data of the new users. The method has the advantages that the method is simple to operate, and convenient to use. The invention provides the dynamic risk customer-group detection algorithm and a corresponding system, which have extremely high commercial values.