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

Description

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.

Steps to reproduce

Clone or download the repository and ensure all required files (code, dataset, and scripts) are available locally. Install the required Python environment and dependencies listed in requirements.txt (e.g., NumPy, Pandas, Scikit-learn, Matplotlib, SHAP).

Institutions

• Southeast University

  Jiangsu, Nanjing

Categories

Licence