Improved GM-SVR combined prediction model of pavement skid resistance condition based on finite data

Abstract

The pavement skid resistance condition of expressway has an important impact on driving safety. With the increase of service life, the pavement skid resistance condition decreases year by year, and an accurate prediction model is of great significance to improve the level of traffic safety. Firstly, the improved GM (1, 1) model and support vector machine regression model are established, and then the two are combined by entropy weight method to obtain the GM-SVR prediction model. In this paper, the skidding resistance index (SRI) of a certain section of Lezi expressway for limited years (2016, 2018 and 2020) is used as the basic data to predict the SRI values in 2021 and 2022. In order to verify the accuracy of the model, the pavement condition index (PCI) of the same road section in 2013, 2015 and 2017 are used as the basic data to predict the PCI values in 2018 and 2019; Taking MAE, MSE and MAPE as test indicators, the predicted values and measured values in 2018 and 2019 are compared and analyzed to test the prediction accuracy.