Hybrid feature selection framework for predicting bridge deck conditions

Bridge decks’ maintenance funding requirements are influenced by bridge decks' current and predicted future conditions. Additionally, the serviceability of bridges may be negatively impacted degradation decks. inspections require considerable effort, time, cost, resources; besides, such introduce hazards safety concerns. This paper introduces a data-driven hybrid feature selection framework for predicting deck deterioration conditions applying it to in Iowa State, USA. Firstly, Boruta algorithm, stepwise regression, multi-layer perceptron employed find best subset features that contribute deterioration. Then, four classification models were developed using features, namely k-nearest neighbours, random forest, artificial neural networks, deep networks. The hyperparameters optimized get their performance. showed comparable performance, forest model outperformed other prediction accuracy with fewer misclassifications. thought reduce field give insights into most influential factors