The purpose of this research is to evaluate and analyze the fitness for human feet in different shoe insoles. We find the most-related sample between foot shapes and insoles by using the grey-relational approach. Based on the plantar pressure measurement on the insoles, we put them into an artificial neural network (ANN) as the training pair (pressure-insole) for network training. After training iterations, the network will have sufficient generalizing capability to classify the pattern of the plantar pressure. Back-propagation neural network (BPNN) is used to convert and classify shoe insoles. By referring to the classified results estimated by the network, a designer can decide on the best direction for the design project. Furthermore, this approach can effectively reduce the design-cycle time and meet the customer’ demands. The results and contributions in this paper are as follows. Firstly, we conducted a foot experiment to verify our research assumptions. Secondly, we investigated the validity of using the grey-relational approach to estimate the fitness of a foot based on the plantar pressure data. Thirdly, we verified the validity of ANN's learning and classifying capabilities. Lastly, we used ANN to learn from the fitness data and predict the most appropriate insoles for the foot.