Condition Prediction of Sanitary Sewer Pipe Data Set with Imbalanced Classification
Karthikeyan Loganathan, Mohammad Najafi, Praveen Kumar Maduri, Kawalpreet Kaur
Galgotias College of Engineering & Technology The University of Texas at Arlington
阅读操作
确认中在文库中上传 PDF 后可生成中文音频讲解。
摘要与影响
Inspection and condition assessment of pipelines play a vital role in the successful operation and maintenance of systems. In the United States, closed-circuit television (CCTV) is the commonly used device for inspecting the inner environment of sanitary sewer pipes. Inspection of every individual sanitary sewer pipe segment is not feasible for any municipality owing to its large inventory of pipes and incurred cost. Machine learning (ML) algorithms, such as logistic regression (LR), k-nearest neighbors (k-NN), and random forests (RF), were employed to develop condition prediction models that could predict the sanitary sewer pipes in need of repair or a maintenance activity. Although the LR model was unable to capture any sewer pipe in poor condition, the same model has resulted in a reasonably higher area under the curve (AUC) value of 0.76. This phenomenon was found to be due to higher imbalance in the data set. Therefore, the study aimed to overcome the limitation of imbalanced classification by employing techniques such as random under sampling and random over sampling. ML algorithms were employed for all three sampled data sets. With an F1-score of 0.94, the RF model outperformed both LR and k-NN models. The developed models can be utilized by utility owners and municipal asset managers to make more informed decisions on future inspections of sewer pipelines.
逐年被引趋势
关键指标
同类平均 = 1
同领域 · 同年份 · 同类型
Google Scholar 与 OpenAlex 的被引统计范围不同,数值存在差异属正常。
AI 辅助阅读
依据:摘要
可就本文提问;依据不足时会说明。
学术脉络
学科主题
工程Water Systems and Optimization
Infrastructure Maintenance and Monitoring · Imbalanced Data Classification Techniques
参考文献 15
此处列出前 3 条
引用本文 2
按被引量排序,此处列出前 3 条