linear, polynomial, RBF and sigmoid

linear, polynomial, RBF and sigmoid. And evaluated its performance using thefollowing parameters: accuracy, recall, precision. Considering overall accuracy,the RBF kernel of SVM outperforms other kernels. They have also tested theclassification accuracy by varying the number of features. And accuracy is higherwith a maximum 13 selected features. In the Kmeans clustering algorithm, theyhave used the unlabeled data with a predetermined number of clusters. Theyhave compared results with supervised and unsupervised models and accordingto the paper, SVM has the highest precision and overall accuracy.Authors of [6] have discussed and concepts of SDN, Network Function Virtualization(NFV), Machine learning, and big data driven network slicing for 5G.In their work, they have proposed an architecture to classify network traffic andused those decisions for network slicing. According to the paper, with the exponentially increasing number of applications entering the network is impossibleto classify traffic by a single classification model. So they have used the Kmeanclustering algorithm to cop this issue. By using this unsupervised algorithm,they have grouped the data set and labeled them. They have set the numberof clusters k=3 associating three bandwidths. With this grouping and labeling,they have trained five classification models: Navie Bayes, SVM, Neural networks,Tree ensemble, Random Forest. And compared its accuracies. The results showthat Tree ensemble and Random forest perform with the same accuracy. Dependon the ML output, bandwidth was assigned in the SDN network applications.They have ed this system by streaming YouTube a video before the classificationprocess and check the quality of the video. And compared it with the quality ofthe video after the classification and bandwidth allocation.In this study, the number of features was selected based on keeping thecompatibility with the implementation (SDN controller) and avoid complexityand heavy computations in the network application. An unsupervised learningalgorithm was used to identify the optimum number of network traffic classesrather than selecting a predefined number of network traffic classes, which makesthis method a more customized network traffic classification solution for networkoperators.