Details

Autor(en) / Beteiligte

• Wang, Changbai
• Xu, Shuzhan
• Liu, Jimin
• Yang, Junxin
• Liu, Chenxu

Titel

Building an improved artificial neural network model based on deeply optimizing the input variables to enhance rutting prediction

Ist Teil von

• Construction & building materials, 2022-09, Vol.348, p.128658, Article 128658

Ort / Verlag

Elsevier Ltd

Erscheinungsjahr

2022

Quelle

Alma/SFX Local Collection

Beschreibungen/Notizen

• •Build a rutting artificial neural network model to improve MEPDG prediction accuracy.•Optimize the rutting prediction by proposed method based on random forest.•Analyze the interaction between variables based on game theory. This paper proposes a method based on a random forest algorithm to optimize the input variables, which successfully improves the prediction accuracy of the rutting artificial neural network (ANN) model. For testing, we collected 5,265 historical observation records in the United States and Canada from the Long-Term Pavement Performance database, including pavement structure and construction, climate, traffic, and performance. Also, we established two additional ANN models to evaluate the model’s performance by using the same data. The results reveal that the proposed model is significantly better than the other two models, and the coefficient of determination R2 and the mean square error MSE are 0.932 and 1.108 mm in the testing set, respectively. As a result, this method can be successfully used in data preprocessing, whereby eliminating insignificant and multicollinear variables needs to be considered carefully.