CNAM November Webinar - Learning from Mistakes: Data Quality and Model Training Challenges in Road Asset Management

Join us for the “Learning from Mistakes: Data Quality and Model Training Challenges in Road Asset Management" webinar on November 24th from 1-2 PM ET.

Many municipalities have been collecting data about their infrastructure and recently have adopted data driven solutions. Machine learning algorithms have become a popular solution for problems in the area of infrastructure management. While the algorithms and codes are almost public and known to everyone, their application is often suboptimal due to some issues that are discussed here. This study presents some of the worst practices in the application of data analytics to infrastructure asset management with emphasis on road asset management. These issues are summarized in three main areas: data retrieval and preparation, data quality, and model training. The paper relies on the pavement performance data retrieved from the Long-Term Pavement Performance (LTPP) with specific focus on two pavement performance indicators: Pavement Condition Index (PCI) and International Roughness Index (IRI). More than 3000 cleaned PCI records and 30,000 IRI records were joined with climatic and operation data were used to demonstrate the best/worst practices. Using this dataset, the study investigated the barriers to implementing data-driven solutions with emphasis on data quality issues (e.g., data scarcity, and representativeness) and model training (e.g., accuracy and overfitting). Flagging erroneous data (anomalies) and limitations of algorithms when applied to different types of data was investigated. The recommendations presented in this study can be used in different sectors of infrastructure management.

Event for CNAM Members and Non-Members

Presenter Bio: 

S. Madeh Piryonesi, PhD, P. Eng: Madeh Piryonesi an assistant professor at the Department of Civil Engineering at Lakehead University. His research is focused on applying data-driven solutions to infrastructure asset management with focus on climate change. In addition to research and teaching, Madeh has also worked in finance and construction industries as a professional and is interested in the intersection of these domains with computer science.

Register Here