This folder contains the code associated with the study: Silva Lopez, R., Baker, J. W., and Poulos, A. (2022). “Deep learning-based retrofitting and seismic risk assessment of road networks.” ASCE Journal of Computing in Civil Engineering, 36(2), 04021038. https://doi.org/10.1061/(ASCE)CP.1943-5487.0001006 The paper presents two main topics. First, it presents the training of a deep neural network. Then, taking advantage of the structure of the neural network, it proposes a retrofitting policy using a modified variable importance algorithm called LIME. Considering that the first part of the study requires training data, and that the size of the files of training data are too big to be attached to GitHub, we will present some example codes to introduce the results of the first section instead of the original codes. The files of this repository are organized in the following way: 1- Trainin_Example.py : Sample script that trains a neural network given a reduced training set. 2- model_fitted_resampled_12732.pkl.zip: .pkl file that contains the final neural network model that predict the traffic performance metric. 4- LIME_Algorithm: Simplified implementation of the modified LIME algorithm that shows as a result a ranking of bridges to be retrofitted. 4- Evaluate_performance.py: Script that evaluates performance of different retrofitting strategies using the neural network. 5- util.py: auxiliary code used to compute network performance There are also two attached folders. "Input" has the files needed to run the experiments. "Figures_Paper" has the codes necessary to run some of the experiments and produce respective figures of the article. The first author Rodrigo Silva-Lopez is open to share all codes and training data upon request to the email rsilval at stanford.edu . Note that more than 38 gb of files are involved in the calculation of this paper. Some files are compressed as .zip due to limits imposed by GitHub. In order for the codes to run properly, you need to unzip those files. Some of these files are in the "input" folder