A data mining method for quality prediction using association rule (DMAR) is presented in this paper.Association rule is used to mine the valuable relations of items among amounts of textile process data for ANN prediction model. DMAR consists of three main steps: setup knowledge data set; data cleaning and converting; find the item set with large supports and generate the expected rules.DMAR effectively improves the precision of prediction in yarn breaking. It rapidly gets rid of the negative influence of training parameters on prediction model. Then more satisfactory quality prediction result can be reached.