Clickstream Content Sessionization

This is an undergraduate thesis for Bachelor in Computer Engineering

The idea of this work is to identify account sharing and classify anonymised user click streams by content, dividing a click stream into sub-sessions and attributing each session to it's current anonymised user.

Useful files

• Thesis presentation
• Jupyter notebook for a brief explanation and the proposed method in action
• LaTex files

The data used for this work was the globo news features from globo dataset, t-SNE was used for dimentionality reduction, affinity propagation for clustering and euclidean distance as threshold.