In the reclamation project in the coastal areas of Fujian，the main soil particles are medium-coarse sand types. In the process of transportation of this type of soil，large resistance is more likely to cause pipe blockage，which will delay the construction process. Under this working condition，it is very important to grasp the real-time information of the conveying resistance in the pipeline. However，under coarse particle conditions，commonly used empirical methods such as the Durand formula method have poor calculation accuracy. In this paper，based on the existing pipeline transportation research and the test data of the actual coastal engineering in Fujian，the Gaussian process regression method and support vector machine method are used to establish the prediction model of the pressure drop in the pipeline. Both regression models can get more ideal effects during the training period，the R2 index of the model is above 0.80. In the prediction period of the model，the R2 index of the support vector machine regression model is 0.78，and the R2 index of the Gaussian process regression prediction model reaches 0.95. The results show that the machine learning prediction model based on Gaussian process regression can better predict the dredging parameters of the medium-coarse sand dredging project.