K-means 군집화 전처리를 활용한 제주도 지방도 포장 소성변형 예측모형 개발

초록

performance for local roads on Jeju Island by applying K-means clustering for data preprocessing, thereby improving the accuracy of the prediction model. Pavement management system (PMS) data from Jeju Island were utilized. K-means clustering was applied, with the optimal K value determined using the elbow method and silhouette score. The Haversine formula was used to calculate the distances between the analysis sections and weather stations, and Delaunay triangulation and inverse distance weighting (IDW) were employed to interpolate the magnitude of the influencing factors. The preprocessed data were then analyzed for correlations between the rutting depth (RD) and influencing factors, and a prediction model was developed through multiple linear regression analysis. The RD prediction model demonstrated the highest performance with an R² of 0.32 and root-mean-square error (RMSE) of 1.48. This indicates that preprocessing based on the RD is more effective for developing an RD prediction model. The study also observed that the lack of pavement-age data in the analysis was a limiting factor for the prediction accuracy. The application of K-means clustering for data preprocessing effectively improved the correlation between the dependent and independent variables, particularly in the RD prediction model. Future research is expected to further enhance the prediction accuracy by including pavement-age data.

키워드