Comparable Analysis on Machine Learning Prediction Models of Mean Annual Ground Temperature in Qinghai Tibet Engineering Corridor

Abstract

Machine learning algorithms have been widely used in the field of permafrost. In order to compare the advantages and disadvantages of different algorithms in the application of mean annual ground temperature, combined with 277 groups of mean annual ground temperature monitoring data along the Qinghai Tibet Highway and Qinghai Tibet Railway, three machine learning models of mean annual ground temperature are established based on mean vegetation index, latitude, elevation, equivalent latitude, surface temperature and ice content: support vector machine, random forest, radial basis function neural network model. Comparing the prediction results of the three models, RBF neural network has the best prediction effect, and the goodness of fit R2 reaches 0.87. The ice content of the prediction factor is an important factor affecting the prediction effect of the model. And compared with previous studies, it shows that the machine learning model is more accurate and stable in predicting mean annual ground temperature.