Prediction of Asphalt Pavement Smoothness Based on Decision Factor and PCA-DBSCAN-RF Model

To enhance the precision of predicting pavement smoothness, this paper presents a model that merges the DBSCAN clustering algorithm with the random forest algorithm. Principal Component Analysis (PCA) is subsequently utilised to attain a reduction in data dimensionality. A DBSCAN clustering algorithm is applied to enable road element clustering. A decision variable ε is proposed which considers the effect of the RQI of neighbouring road sections. Finally, a distinct random forest model is constructed for each cluster of data and one consolidated model is created for all data. The results show that the prediction accuracy of the model with ε is better than that of the regression model, the random forest model, and the combined model without ε. The reduction in MAE and RMSE ranges from 6% to 78% and 0.3% to 59%, respectively. Modelling can offer guidance for the upkeep of road surfaces.