Abstract: The over-the-horizon propagation characteristics of evaporation duct is an important factor affecting the marine radio system. Therefore, accurate prediction of evaporation duct is the basis of system evaluation. In this thesis, an evaporative duct inversion method based on sparse sampling propagation data and Light Gradient Boosting Machine (LightGBM) of decision tree algorithm is proposed. Firstly, the training set and the test set are constructed by the sparse sampling propagation data obtained by the parabolic equation. Secondly, the decision tree LightGBM algorithm is used to build the inversion model, and the accuracy of the model is improved by adjusting the parameters to achieve higher inversion accuracy. Finally, the trained Light- GBM model is used for evaporation duct inversion, and the probability distribution of inversion results is analyzed. The result shows that the evaporation duct inversion method based on sparse sampling propagation data can effectively and quickly realize the evaporation duct inversion, which provides a new way for the prediction of evaporation duct at sea.