Optimizing SVM models as predicting tools for sewer pipes conditions in the two main cities in Colombia for different sewer asset management purposes

Currently, sewer utility systems require extending management activities by developing tools, such as deterioration models, to face their aging problem. In the literature on sewer asset management, Support Vector Machines (SVM) have been a useful tool to predict and forecast pipe's structural conditions. In this paper, the differential evolution method was implemented as an optimization tool for the hyper-parameters combinations to use in SVM-based models for two different management objectives (network and pipe levels). These models were applied to Colombia's main cities of Bogota and Medellin, resulting in a less than 6% deviation in the prediction of structural conditions in both cities at a network level. The DE-optimized SVM models at the pipe level show higher percentages of correct predictions in all structural conditions than non-optimized SVM models (conventional SVM model) for specific management objectives. Therefore, the relevance of optimizing the hyper-parameters of SVM models to improve predictions of the structural condition of unspecified sewer assets became apparent.

Automated summary

By optimizing hyper-parameters, SVM models in sewer asset management have higher predictive accuracy, and the differential evolution method can be used to enhance model performance.