Evaluation and Prediction of Pavement Deflection Parameters Based on Machine Learning Methods

The deflection measurements made using Falling Weight Deflectometers (FWDs) are widely used in the back-calculation of pavement layer moduli. Pavement structural characteristics, changes temperature, and other related factors exert a significant effect on measurements. Therefore, three machine learning methods—Classification Regression Tree (CART), Random Forest (RF), Gradient Boosting Decision (GBDT)—were to evaluate importance influencing including FWD test conditions, parameters, climatic factors, traffic level, rehabilitation service age, basin this study. results indicated that number was an important feature for all but its lg(D0–D20) lg(D0–D30) smaller than relative asphalt layer, base, subbase > base; their lg(D20–D60), lg(D30–D60), lg(D30–D90) base subbase; lg(D90−D120) lg(D60–D120) layer. Among condition variables, drop load most factor second-layer temperature also lg(D0–D20), lg(D0–D30), lg(D0–D45). precipitation greater freeze index. prediction shown accuracy GBDT as high 99%. Besides, outperformed RF, RF CART. analyses between parameters especially characteristics pavement, provide theoretical evidence evaluation strength basis data.