Traffic delays at the United States-Canada border crossings have adverse effects on the economy as well as the environment. This thesis aims to predict passenger cars’ traffic delays at the three Niagara Frontier Border Crossings, namely the Peace Bridge, the Lewiston-Queenston Bridge, and the Rainbow Bridge for the next 60 minutes into the future, using border wait time data collected by Bluetooth readers recently installed at the crossings. Firstly, the delay data were analyzed to assess the influence of bridges, weekdays and weekends, months/seasons, holidays, and direction of travel on traffic delay. After which, traffic delays were predicted using four deep learning techniques: Multilayer Perceptron (MLP), Convolutional Neural Networks (CNN), Long Short-Term Memory Recurrent Neural Networks (LSTM-RNN), and Gated Recurrent Unit Recurrent Neural Networks (GRU-RNN). The prediction accuracies of these models were evaluated by computing the Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and R-squared. The study also compared the performance of models trained on the whole data set (called ‘complete set’ models), versus models that were developed separately for weekdays and for weekends, to examine the effect of data classification on the models’ predictive accuracy. The results suggest high-level accuracy of the deep learning techniques in predicting future traffic delays at the border crossings, with MAEs less than 3.5 minutes in predicting delays for up to 60 minutes into the future. However, no one deep learning technique emerged as a clear winner among others in predicting the delays. Findings of this thesis can guide future travelers in choosing the border crossing with the least delay, which in turn, should help in increasing the efficiency of the borders.