Numeric Prediction Algorithms for Bridge Corrosion
The research reported in this article was conducted to mainly explore the two common numeric prediction techniques, the model tree and the regression tree, when used in conjunction with bagging as a wrapper method. Bagging is used to improve the prediction accuracy of these two algorithms, and results are compared with the ones obtained earlier by the k-nearest neighbor (KNN) algorithm. From the conducted experiments, both the bagged regression tree and bagged model tree produce better results than not only their corresponding regression tree and model tree alone, but also the KNN with optimal value of k equal to 7. In addition, the bagged model tree yields the lowest prediction errors and a highest correlation coefficient of 0.81. It is demonstrated that it is feasible to use the bagged model tree for engineering applications in prediction problems such as estimating the remaining service life of bridge decks.
Use and reproduction:
All rights reserved