Network congestion in the heterogeneous Internet, which is connected by millions of asynchronous systems, poses a serious threat to communication and intermediate nodes that falls under its path. The primary cause of network congestion is that data in networks are overloaded and available resources are inadequate to contain such traffic loads. An enormous amount of proposed approach towards network congestion is based on conventional control methods in the form of mathematical and linear models. However, the explosive growth of the Internet, its traffic and diversification of network applications has limited conventional control mechanism from scaling up and providing an effective solution. Although conventional congestion methods improve the level of control, the vulnerability of linearisation and varying network parameters makes it difficult to provide an efficient solution. In this paper, the problem of congestion is addressed via exploring computational intelligence (CI) methodology and proposing a fuzzy inference engine for congestion control in network edge and bottleneck link environments. Furthermore, through extensive simulation experiments, the results demonstrate that the proposed CI method improves network edge performance during congestion prior to conventional control methods.