Detection of water pipeline leakage based on random forest

Abstract

Pipeline leakage is a great concern for the transportation industries and researchers have been devoted in leakage detection for a long time. Machine learning is developed for leakage recognition recently and it can help to achieve the leakage detection. However, the effect is limited by feature complexity and noise. As a machine learning method, Random Forest (RF) is good at handling with high-dimensional data and predicts well even when the signal is interrupted by noise. As a result, RF was applied to better deal with the leakage detection. Researches herein have compared the RF classifier and other well-developed machine learning methods in respects of the classification accuracy and calculation time. The result indicated that the RF classifier outperformed Support Vector Machine (SVM), Artificial Neural Network (ANN), k-Nearest Neighbors (k-NN) and Decision Tree (DT) classifiers, with the classification accuracy of 88.33%.