This paper reports on the construction of an integrated tool consisting of a neural network and subjoined local approximation technique for application to the sewing process, especially for selecting optimal interlinings for woolen fabrics. A single hidden layer neural network is constructed with five input nodes, ten hidden nodes, and two output nodes. To train the network with a back-propagation learning algorithm, the mechanical parameters used as inputs for the fabrics are tensile energy, bending rigidity, bending hysteresis, shear stiffness, and shear hysteresis, while mechanical parameters used as outputs for the interlinings are bending rigidity and shear stiffness, all of which are measured on the KES- FB system. Even though the back-propagation algorithm has a higher learning accuracy and can be successfully used to select the appropriate interlining, its learning process is too slow and it gets stuck in a local minimum. This research presents a few methods for improving the efficiency of the learning process. The raw data from the KES-FB system are nonlinearly normalized, and input orders are randomized. This proce dure produces a good result, such that the error values of the prediction model are low despite the relatively small data set for training. After training, the optimal interlinings for unknown fabrics can be correctly selected through mapping and the local approximation method.