Human trajectory data and internet traffic mining using improved multi-context trajectory embedding service usage classification model

Due to the rapid growth of mobile messaging Apps, the classification of Internet traffic into different types of service usages has become a vital process to handle the location-based social networks. In previous researches, Improved Multi-Context Trajectory Embedding Model (IMC-TEM) was proposed to analyze and mine the human trajectory data using multiple context information of trajectory data. However, this model does not consider Internet traffic classification that investigates how to use encrypted Internet traffic for classifying service usages. Therefore in this paper, IMC-TEM is incorporated with CUMMA model to classify the service usage using both Internet traffic data and contextual information of trajectory data generated by messaging Apps. In this model, four major processes are performed to predict the service usages and end-user behaviors efficiently. Initially, traffic segmentation process is performed based on the hierarchical clustering with threshold heuristics that segments the Internet traffic into sessions and dialogs. After that, features are extracted from the segmented traffic based on the packet length and time delay. Then, Random Forest (RF) classifier is applied to classify the service usage types. Moreover, clustering-Hidden Markov Model (HMM) is introduced to detect mixed dialogs from outliers and decompose those into sub-dialogs of single-type usage. Finally, the performance effectiveness of the proposed model is evaluated through the experimental results using different real-world datasets.