Network user behavior analysis using machine learning

• Main Authors: Borda, Jose Ruel M., Encarnacion, Justin P., Go, Jan Carlo L.
• Format: text
• Language: English
• Published: Animo Repository 2015
• Online Access: https://animorepository.dlsu.edu.ph/etd_bachelors/11838
• Institution: De La Salle University

Bandwidth Management is the process of measuring and controlling the communications on a network to avoid overtaxing the network, which could result in network congestion and poor performance. Bandwidth Management is becoming an essential part in network management. Current bandwidth management solutions that are available in the market are static in nature. These solutions lead to an issue where the bandwidth itself is not being properly allocated to the user's requirements. Machine Learning is a method of teaching computers to make and improve predictions or behaviors based on existing or collected data. The study uses machine learning algorithms to learn the different habits of users on the network, in order to know how to efficiently allocate the required bandwidth for the users based on those habits. A network packet analyzer tool, Wireshark, is used to collect packet information for the use of network analysis. The collected data is then parsed in a way to be usable for construction of a machine learning model through the usage of WEKA. Each model is then verified and tested with various metrics such as kappa coefficient, F-measure and precision. The model with the best results based on those metrics, in this case J48 decision tree algorithm, can then be said to be the most efficient machine learning algorithm that can be used for this dataset. The model generated reflected the network usage behavior of the data collected. This model can then be used in a practical application such a bandwidth management.