This paper is conducted to mitigate the traffic congestion using Machine Learning (ML) approaches. It is aimed to mitigate the traffic congestion in Kingdom of Bahrain by predicting the solution needed from construction and traffic engineering perspective. This paper used a primary data, extracted from the reports and documents provided by the Ministry of Works who is responsible of the roads’ construction in Bahrain. The extracted dataset has imbalanced classed; therefore, three oversampling methods were applied on the dataset; Synthetic Minority Oversampling Technique (SMOTE), Adaptive Synthetic (ADASYN), Random Oversampling (ROS) and the fourth one is the Hybrid and the combination of the three datasets. The models are built using five ML approaches; Decision Tree, Random Forest, k-Nearest Neighbors, Support Vector Machine and Multilayer Perceptron. The Hybrid dataset produced the high accuracy among all the ML models. K-Nearest Neighbors and Random Forest resulted the best accuracy 100% and 99.9%. This paper proved that the ML approaches are empowering the decision making in taking the suitable solution for road construction to mitigate the traffic congestion.