User Behavioral Modeling for Web-based Systems

• Author / Creator

  Sajjadi Ghaemmaghami, Saeedehsadat

• Today’s users are spending more time on web applications. Many users browse
  web applications and navigate through different web pages. They may have
  different interests, especially when it comes to large-scale applications. The more
  the developers of the applications know about their users’ needs and interests, the
  smarter choices they will make for their application’s development. Inferring a
  behavioral model from users’ navigation patterns in a web application helps
  application providers to understand their users’ interests. A navigational pattern is
  a record of where a user visits; the pattern is extracted from the start to the end of
  a user session. User navigation information is obtained by collecting the data in a
  log file.
  Some studies instrumented the application’s web pages to collect data and then
  model user navigational behaviors. To instrument a web page, the source code of
  the program is modified with additional commands. However, this can be difficult
  when the source code is inaccessible. Ideally, a user behavioral inference process
  should not be required to instrument the application’s web pages to generate a user
  behavioral model.
  Also, a model generation approach needs to support the evolution of web
  applications. A behavioral model should be generated incrementally during its
  evolution and should play a role in the application’s evolution (upgrading)
  procedure. This can help sustain web applications.
  Inferring a model by predicting and analyzing users’ navigational behaviors isiii
  necessary to understand users’ interests. Developers can identify interesting (from
  users’ perspective) or problematic pages of applications and therefore improve the
  application design. Analyzing the behavioral model helps to detect design
  anomalies such as dead-ends; pages in which users are being prevented from
  leaving the page without closing it. Detecting dead-ends can significantly help in
  addressing design anomalies and providing solutions to retain users. Satisfied
  customers are more likely to stay with the company and contribute to its success.
  It is ideal to analyze web pages to model user behaviors. Web page analysis
  methods utilize web page segmentation which is the process of segmenting a web
  page into different blocks, where each block contains similar components in terms
  of structural, visual, or contextual similarity. Current segmentation methods use
  the Document Object Model (DOM) structure of a web page and vision-based
  techniques to segment a web page. However, current methods do not consider
  semantic analysis to categorize pages. Semantic analysis includes extracting text
  from segmented blocks, computing textual similarity, and regrouping blocks.
  In this research, we attempt to bridge several gaps in all the above-mentioned
  areas. Firstly, we provide an automated approach, with no instrumentation, to
  generate user behavioral models. We evaluate the utility of our approach by using
  it on a large-scale mobile and desktop application. Also, we evaluate the evolving
  properties of interaction patterns against the inferred behavioral models using an
  analysis engine.
  Next, we present a new combination model of web page segmentation, namelyiv
  Fusion-Block, by dividing the content of a web page into blocks by initially
  considering human perception (inspired by Gestalt laws of grouping) and
  subsequentially re-segmenting initial similar blocks using semantic text similarity.
  Hence, in the next part of our research, we improve the segmentation model,
  namely Integrated-Block, by merging the DOM structure, vision-based, and textbased similarity metrics of web pages. Finally, to verify the effectiveness of our
  approach, we applied it to the public datasets and compared it with the five existing
  state-of-the-art algorithms. We demonstrate the value and novelty of the presented
  solutions using extensive evaluations throughout the thesis.

• Subjects / Keywords

  • User behavioral model
  • Reinforcement learning
  • web page segmentation
  • semantic text similarity

• Graduation date

  Spring 2022

• Type of Item

  Thesis

• Degree

  Doctor of Philosophy

• DOI

  https://doi.org/10.7939/r3-04x3-vt85

• License

  This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.