Wheat is the main food crop in the world. The planting of winter wheat in alpine regions can not only improve the land use efficiency, but also increase the income of farmers. The planting of winter wheat in alpine regions depends on whether it can survive the winter smoothly. Therefore, the study on the physiological process of winter wheat has application value.Over a three-year field experiment, the relationships between osmotic and anti-oxidative stress in tillering nodes and winter wheat overwintering capacity in alpine regions was examined. A multi-index cold-resistant overwintering capacity model of winter wheat was established using a (machine learning) logistic regression algorithm. The experiment included three winter wheat varieties (Dn1, Dn2, and J22) and eight key physiological indicators of cold resistance. The results showed that in terms of osmotic stress, the water content of tillering node was about 70%, and the Pro content was 2000-2500 μg g-1, soluble sugar content of 1000 μg g-1 above is more conducive to winter wheat overwintering; In terms of antioxidant stress ability, all indexes were maintained at a high level, which worked together to maintain the stability of antioxidant system. Through the analysis of receiver operating characteristic (ROC) curve of cold resistance of each index, six indexes of tillering node water content, hardness, pro content, soluble sugar content, MDA content and POD activity were selected to establish the cold resistance model of winter wheat overwintering, and the effect of the model was evaluated by ROC curve.The winter wheat overwintering model established in this study can provide a convenient and quick method for evaluating the overwintering capacity of novel winter wheat varieties.