In current scenarios, a variety of network applications are being developed for meeting the growing business needs for Industries. These applications are complex in nature, use new protocols and port numbers from the range of the dynamic ports. For Quality of Service (QoS) assessment of these applications, traffic classification is the first and most important step. Each network application has distinct characteristics and its generated traffic also implicitly follows these characteristics. The main focus is on collecting the set of those traffic statistical features which are most suitable to correctly represent the applications under consideration for the classification task. In this paper, focus is on traffic classification in the Software Defined Networks (SDN). This paper has a proposed traffic classification and QoS assessment architecture. This proposed architecture is used for classifying the incoming traffic at the network edge devices. For classification purpose, six different machine learning algorithms have been used. Various type of traffic have been generated by using Distributed-Internet Traffic Generator. Further, the performance of these classifiers have been evaluated using the confusion matrix. A comparative performance study is done in the last. This work is based on the traffic statistical feature analysis and collected statistical features are evaluated for finding their relative importance in the task of classification. The proposed architecture works in the existing SDN environment and requires low computation, as only the important features have been passed and analyzed for the traffic classification.