A Pipeline Blockage Identification Model Learning from Unbalanced Datasets Based on Random Forest

Aiming at the problem of decreased accuracy of operating state recognition caused by the unbalanced data acquisition between the normal and blocked failure states of urban drainage pipelines, a new method of pipeline blockage state recognition based on unbalanced data is proposed. The experimental results show that the random forest algorithm, which uses bootstrap sampling and simple voting methods to integrate decision trees, has a good effect on the pipeline blocking state recognition of unbalanced data.