Urban traffic congestion represents a complex challenge influenced by many dynamic factors. Peak periods typically exacerbate congestion, while bad weather can slow vehicle movements and increase travel times. Accidents and road closures cause sudden and unexpected disruptions, making traffic management a constant challenge. Using a dataset of over 66,000 GTFS records with machine learning classifiers like Random Forest, XGBoost, CatBoost, and Decision Tree models, the study seeks to forecast traffic conditions. SMOTE is used to ensure greater representation of minority classes in order to solve the dataset's intrinsic imbalance, and feature scaling enhances model convergence. With an accuracy of 98.8%, Random Forest was the most accurate model for this challenge. The outcomes demonstrate that the system is able to precisely forecast traffic in real-time, which aids in route planning, traffic control, and enhancing urban mobility.