[Paper Review] Memory and Structure in Human Navigation Patterns.

Philipp Singer, Denis Helić+2 more|arXiv (Cornell University)|Feb 4, 2014

Web Data Mining and Analysis24 references9 citations

TL;DR

This paper evaluates the appropriateness of first-order vs. higher-order Markov chain models for modeling human web navigation. Using advanced inference methods on two navigational datasets, it finds that while memoryless models suffice at the page level, topical-level navigation reveals memory effects, necessitating higher-order models and highlighting the need for context-aware navigation modeling.

ABSTRACT

One of the most frequently used models for understanding human navigation on the Web is the Markov chain model, where Web pages are represented as states and hyperlinks as probabilities of navigating from one page to another. Predominantly, human navigation on the Web has been thought to satisfy the memoryless Markov property stating that the next page a user visits only depends on her current page and not on previously visited ones. This idea has found its way in numerous applications such as Google's PageRank algorithm and others. Recently, new studies suggested that human navigation may better be modeled using higher order Markov chain models, i.e., the next page depends on a longer history of past clicks. Yet, this finding is preliminary and does not account for the higher complexity of higher order Markov chain models which is why the memoryless model is still widely used. In this work we thoroughly present a diverse array of advanced inference methods for determining the appropriate Markov chain order. We highlight strengths and weaknesses of each method and apply them for investigating memory and structure of human navigation on the Web. Our experiments reveal that the complexity of higher order models grows faster than their utility, and thus we confirm that the memoryless model represents a quite practical model for human navigation on a page level. However, when we expand our analysis to a topical level, where we abstract away from specific page transitions to transitions between topics, we find that the memoryless assumption is violated and specific regularities can be observed. We report results from experiments with two types of navigational datasets (goal-oriented vs. free form) and observe interesting structural differences that make a strong argument for more contextual studies of human navigation in future work.

Motivation & Objective

• To assess whether human navigation on the Web adheres to the memoryless Markov property or requires higher-order models.
• To compare the complexity and utility of higher-order Markov chains against first-order models in modeling real-world navigation.
• To investigate structural differences in navigation patterns between goal-oriented and free-form browsing behaviors.
• To identify when and why the memoryless assumption breaks down, particularly when abstracting navigation to the topical level.
• To advocate for context-sensitive modeling of human navigation, especially at the topic level where memory effects emerge.

Proposed method

• Applies a diverse set of advanced inference techniques to determine the optimal Markov chain order for given navigation data.
• Uses two distinct datasets—goal-oriented and free-form navigation—to evaluate model performance across different browsing contexts.
• Abstracts page-level transitions into topic-level transitions to assess memory effects beyond individual pages.
• Employs statistical inference methods to compare model complexity against predictive utility in higher-order Markov chains.
• Analyzes structural regularities in navigation sequences to detect violations of the memoryless assumption.
• Evaluates trade-offs between model complexity and accuracy to determine practical applicability of higher-order models.

Experimental results

Research questions

• RQ1 Does human web navigation satisfy the memoryless Markov property at the page level?
• RQ2 How does the complexity of higher-order Markov models compare to their predictive gain in modeling navigation?
• RQ3 Are there structural patterns in navigation that violate the first-order Markov assumption when viewed at the topical level?
• RQ4 How do goal-oriented and free-form navigation behaviors differ in terms of memory and structural regularity?
• RQ5 Under what conditions does the memoryless assumption break down, and what implications does this have for navigation modeling?

Key findings

• The memoryless first-order Markov model remains a practical and sufficient model for human navigation at the individual page level.
• Higher-order Markov models exhibit faster-growing complexity relative to their utility, making them less efficient for page-level modeling.
• At the topical level, the memoryless assumption is violated, revealing consistent structural patterns in navigation behavior.
• Goal-oriented and free-form navigation exhibit distinct structural regularities, suggesting context-dependent navigation dynamics.
• The findings support the need for more contextual and topic-aware models in future navigation research, especially beyond page-level analysis.
• The study provides empirical evidence that higher-order models are necessary when abstracting navigation to topics, even if less so at the page level.