Bayesian Optimization-Based SVR and RF Models for Predicting Compaction Quality of SBS-Modified Asphalt Pavements Using Intelligent Compaction Data

This repository contains the source code for the study titled “Bayesian Optimization-Based SVR and RF Models for Predicting Compaction Quality of SBS-Modified Asphalt Pavements Using Intelligent Compaction Data.” The code implements Bayesian Optimization (BO)-tuned Support Vector Regression (SVR) and Random Forest (RF) models for predicting Intelligent Compaction Measurement Values (ICMVs) and Non-Nuclear Density Gauge (NNDG) values using field-collected data from SBS-modified asphalt pavement projects. Input variables include section length, vibratory roller passes, roller speed, vibration amplitude, and mat temperature. The framework also includes multi-output modeling, model evaluation, and SHAP-based explainable AI analysis to quantify feature importance and interpret compaction behavior.