DONGYiguang,WANGDan,HUANGMengni,et al.Risk Prediction of Drainage Pipelines Based on Logistic Regression Model[J].China Water & Wastewater,2026,42(12):71-78.
Risk Prediction of Drainage Pipelines Based on Logistic Regression Model
China Water & Wastewater[ISSN:1000-4062/CN:12-1073/TU]
volume:
第42卷
Number:
第12期
Page:
71-78
Column:
Date of publication:
2026-06-17
- Keywords:
- drainage pipeline; risk prediction; Logistic regression; pipeline defect; geographic information system (GIS)
- Abstract:
- In recent years, frequent issues in underground drainage networks have posed serious threats to aquatic environments and public assets. Conducting risk prediction for drainage networks holds significant importance for optimizing pipeline maintenance and rehabilitation plans, as well as enhancing operational safety. The complete detection and maintenance data of underground pipelines in Shenzhen provide a robust foundation for precise risk prediction. Meanwhile, advancements in geographic information system (GIS) and machine learning offer new research perspectives for pipeline risk assessment. This study utilizes measured pipeline data from a district in China, selecting 80% of risk points as training data. Seven risk factors include sewage/stormwater pipeline type, pipe age, diameter, length, material, elevation, and burial depth. These factors are incorporated into a binary Logistic regression model for risk prediction. Model performance is evaluated using a confusion matrix and ROC curve. The results demonstrate: ① Sewage pipelines, aging pipes, and smaller-diameter pipes exhibit significantly high-risk characteristics; ② The model achieves an overall accuracy of 79.0%, with AUC values of 80.1% and 84.6% for the training and validation sets, respectively, indicating strong predictive capability; ③ Pipe segments with risk probability grades Ⅲ and Ⅳ accounted for 27.33% of cases, covering 61.90% of risk points in the study area, which aligns with the actual risk distribution. This research confirms that the risk prediction method, integrating multi-dimensional risk factors and Logistic regression modeling, effectively evaluates the risk distribution of drainage networks in the study area, providing reliable support for scientific management and decision-making in drainage pipeline maintenance.
Last Update:
2026-06-17