Moving from legacy internet applications, such as web, and e-mail, to dynamic complex applications, such as video streaming, and file sharing, improves the underlying network architectures to provide new services with high QoS requirements. Thus, a new network paradigm is developed and deployed; Software Defined Networking (SDN). As a result of the significant impact of using the internet, network traffic is growing up, and the network itself is becoming more overloaded. Therefore, new tools that can cover new requirements for high-quality services are becoming mandatory. Inherently, traffic classification (TC) has gained continuous interest as an important decision-centric approach to deliver the quality of service. The current main challenge is how to classify flows efficiently. In this work, a new method of classification for incoming flows is proposed. It is based on Deep learning and consists of using the MultiLayer Perceptron Model (MLP) to classify flows according to its constraints throughput and delay. Experimental results prove that the proposed approach outperforms parallel solutions in terms of precision. It can classify traffic with more than 97.7% precision compared to the Linear Regression classification and the Fuzzy Decision-tree model.