This brief presents an innovative algorithm integrated with particle swarm optimization and artificial neural networks to develop short-term traffic flow predictors, which are intended to provide traffic flow forecasting information for traffic management in order to reduce traffic congestion and improve mobility of transportation. The proposed algorithm aims to address the issues of development of short-term traffic flow predictors which have not been addressed fully in the current literature namely that: 1) strongly non-linear characteristics are unavoidable in traffic flow data; 2) memory space for implementation of short-term traffic flow predictors is limited; 3) specification of model structures for short-term traffic flow predictors which do not involve trial and error methods based on human expertise; and 4) adaptation to newly-captured, traffic flow data is required. The proposed algorithm was applied to forecast traffic flow conditions on a section of freeway in Western Australia, whose traffic flow information is newly-captured. These results clearly demonstrate the effectiveness of using the proposed algorithm for real-time traffic flow forecasting.