Sponsored Research (Electronics and Telecommunication Research Institute)
Research Motivation
Leaking water pipes can result in various unintended consequences, such as property damage, environmental contamination, and resource wastage, potentially leading to increased indirect expenses and adverse impacts. In response to this concern, the ETRI has made significant efforts to develop a decision support system designed to systematically manage issues related to water leaks.
Key Idea
This research proposes a methodology that consists of three different modules: 1) the data pre-processing module that implements a data pipeline tailored for handling water leak associated datasets, 2) the classification modeling module that employs machine learning and deep learning algorithms to determine class membership, and 3) the model evaluation module that utilizes the concept of a confusion matrix to assess the performance of the model, including metrics such as precision and recall.
Overview of the proposed methodology
Results
The applicability of the methodology was demonstrated through a validation study using real-world operational data, specifically within the city of Daegu. The validation results showed an overall accuracy of approximately 96% for the validation datasets, indicating that the proposed methodology (i.e., line-based classification) could enhance the fidelity of the system as it demonstrated a greater capability to accurately identify class membership compared to the previous approach (i.e., point-based classification), which achieved an accuracy of approximately 80%.
Publications
[1] ETRI journal (Under review)
コメント