Network-Traffic-Classification
Network Traffic Classification is a machine learning project for classifying network flows captured from a Docker-based SDN lab environment. The dataset includes over 300,000 flows analyzed with nDPI, with more than 100 application protocols grouped into 10 traffic classes such as Web, Streaming, Social Media, and DNS. Seven flow features are used for training, spanning protocol metadata, packet counts, and byte counts. The repository supports reproducible experiments across multiple supervised algorithms. Reported results include KNN at 97.24%, Random Forest at 96.69%, Decision Tree at 95.80%, and a custom PAA method achieving 99.29% accuracy. It provides a modern modular Python package with a command-line interface for training and evaluation, alongside original per-algorithm folders containing legacy scripts, notebooks, saved models, and confusion matrices. DNN experiments using TensorFlow are supported through optional dependencies. Typical workflows include training classifiers on the provided CSV datase