To address the daily coarse release control problem in semiconductor Assembly and Test Manufacturing(ATM) system, a novel release control strategy based on product cluster analysis was proposed to reduce conversion cost in ATM. A new quantity limited and feature weighted K-modes clustering algorithm was used to cluster the products in weekly scheduling, in which the number of bottleneck capacity types was took as cluster number and the number of points in each cluster class was limited. According to the attribute information of conversion cost, the cluster analysis was made. Based on the results of cluster analysis, daily release products and product-quantity were determined by the integrated strategy of product average distribution strategy and product-quantity average distribution strategy. The effectiveness and advantage of proposed strategy was proved by experiments.