An accurate short-term traffic forecasting model serves as an integral part to enhance the efficiency of vehicle rerouting and traffic light control strategies. The information exchange (pheromone) behavior of ants has been applied to forecast traffic conditions in existing pheromone models. These models were developed to forecast congestion on roads with signalized intersections by considering only green and red phases. Motivated by this issue, three short-term traffic forecasting models are proposed:(i) Extended Pheromone Model (EPM),(ii) Extended Pheromone Model with epsilon-Support Vector Regression (ε SVR-EPM), and (iii) Extended Pheromone Model with Artificial Neural Network and Particle Swarm Optimization (ANNPSO-EPM). It is worth noticing that EPM is an algorithmic model whereas the other two are machine learning models. In all proposed models, a new color pheromone concept is proposed with two significant contributions. First, the color pheromone concept is developed to capture stochastic traffic conditions on the roads with non-signalized intersections. Second, the proposed concept is further extended to include all three color phases (red, yellow and green) to forecast dynamic changing traffic behaviors for roads with signalized intersections. The proposed color pheromone concept in EPM, ε SVR-EPM, and ANNPSO-EPM is different from the existing models as it dynamically switches its computation techniques based on traffic light phases. All three proposed models can be realized through a Pheromone-based Multi-Agent System composed of Vehicle Agents and Intersection Agents coordinating locally with one another To promote practicality, Singapore City Hall map is employed in a microscopic simulator of Simulation of Urban Mobility (SUMO), showing that all proposed models outperform the other existing pheromone models.