Seam pucker is an important factor to evaluate the quality of the garment product. The mechanical and structural parameters of the fabric are one of the groups of factors which influence significantly on seam pucker. Establishing a seam pucker predicting model with acceptable cost, accuracy and friendly use is very necessary for the garment industry. In this article, k- Nearest Neighbor Algorithm is used to develop predicting model of the seam pucker. The predictions are based on fabric properties. Mechanical and surface properties of fabric were measured in the low-stress region on the Kawabata Evaluation' System for Fabrics (KESF), and structure properties were calculated by the Peirce model. Four methods (simple kNN, Greedy Forward Search, Backward Elimination and Gradient Descent) were applied for predicting seam pucker to determine optimal weights. The results indicate that designed model is effective for predicting warp and weft seam pucker grades. Furthermore, there is a significant correlation between seam pucker predicted from the model and actual seam pucker grades evaluated by AA Tee 88B standard. The results also show that, Gradient Descent is the most suitable method among those applied to determine the weights that best match with the predictive model for seam pucker.