Machine Learning User Web Interactions

Christopher Wei

ABSTRACT

Testing web services prior to launching is often difficult because
design elements will affect the way users interact with content, which
may result in behaviors different than in- tended. Besides potentially
harming the efficacy of such a website, this also results in web
browsing patterns that may negatively impact the performance of the
server. To better simulate the browsing patterns of a user, we propose
using a machine learning algorithm to predict the browsing patterns of
a user.  This paper will explain the overall design, implementation,
and analysis of user behavior on websites, starting with a data
collection browser extension which will provide the necessary feature
sets to train on, feeding into a training algorithm design that
accommodates the feature set, culminating in a tool that can predict
the browsing patterns of a user given any webpage.