Technology-enabled snow and ice sports data optimization facilitates the development of snow and ice sports training strategies and provides support for ensuring the safety and enhancing the training effect of youth snow and ice sports. This paper designs a system to collect and track snow and ice sports data. It is designed to extract electromyographic signal data from snow and ice sports training through a high-frequency wavelet function. A clustering algorithm is used to process and optimize snow and ice sports training data, and then a sports feature data mining model is established for snow and ice sports auxiliary decision support. The deep learning TensorFlow framework was used as the basis to realize the construction of a generalized dynamic fuzzy neural network, and a variable sliding window was used for model learning in order to realize the accurate prediction of the training load of youth snow and ice sports. The GD-FNN model predicted the muscular strength level of snow and ice sports training with a relative error of only 1.74%, and it also accurately predicted the blood oxygen saturation. The RMSE value of blood oxygen saturation was only 0.68. Under the personalized cyclic impedance training mode, the hemoglobin content of the experimental group was 152.64±8.76 g/L, which was significantly different from that of the control group at the 1% level. Relying on the training data of youth ice and snow sports in the Belt and Road, clarifying the physiological conditions of athletes in different countries can help to enhance the synergistic improvement of the ice and snow sports level in the Belt and Road countries.