Concrete Deck Performance Model based on Hybridization of Ensemble Learning and Artificial Neural Network

Abstract

 The main aim of the bridge maintenance  strategy is to evaluate the performance based on their past  conditions and taking appropriate actions for future  decisions. Further, taking the appropriate actions at right  time reduces the maintenance cost of the bridges. Earlier,  man-power-based methods are used for bridge  maintenance strategy. However, this method is time  consuming, human error while predicting the conditions,  and more cost demandable. To overcome these issues,  machine learning models-based bridge maintenance  strategy are designed in which machine learning models  are trained with past conditions of the bridges and future  conditions are predict. Thus, this paper proposes a concrete  deck performance model based on the hybridization of  machine learning algorithm. The main contribution of this  work is to enhance the performance of the model by hybrid  the prediction of machine learning algorithms. This is  achieved by ensemble learning and artificial neural  network algorithm. Besides that, feature selection algorithm is applied in the pre-processing of the model. To  validate the performance of the proposed model, standard  national bridge inventory (NBI) database was taken under  consideration and various performance metrics are  measured for it. The result shows the proposed model  achieves high value of these performance metrics such as  accuracy (≅ 0.90207), recall (≅ 0.9363), precision (≅ 0.8855), and F-score (≅ 0.92935) and performance  superior over autoencoder and random forest algorithm.