Data Cleaning Method for Asphalt Pavement Maintenance Decision-Making Based on ANN and Reliability

Currently, the quality of the data in most pavement management systems is poor, which affects the accuracy of pavement performance evaluation and the reliability of maintenance decision-making. This study proposed a cleaning method for the data related to asphalt pavement maintenance decision-making. Firstly, an analysis of pavement management data of highways in Shanxi Province was conducted. Then, a pavement maintenance engineering data filling model based on an artificial neural network was constructed. Compared to the logistic regression model, the accuracy of prediction has improved by 17.63%. A cleaning rule that fixed the anomaly in the pavement condition data was then proposed. The rule was based on the quartile spacing method and reliability analysis. The method proposed in this study considers and cleans the pavement condition data and maintenance engineering data simultaneously. In this way, the overall data quality can be improved, and a more reliable maintenance decision program can be presented.