Proposing a New Structure for Web Mining and Personalizing Web Pages

Hamid Alinejad‐Rokny, Mostafa Keikhay Farzaneh, Amir Goran Orimi, Mir Mohsen Pedram, Hojjat Ahangari KiasariJournal of Emerging Technologies in Web Intelligence 2013 · DOI

TL;DR

This paper proposes a novel, flexible structure for web mining and personalization, specifically tailored for e-commerce websites. It aims to enhance user experience and potentially increase profits by dynamically tailoring web pages and recommendations based on individual user preferences and behavior.

What Problem Does This Paper Address?

The internet has become a vast repository of information, but navigating and understanding its content remains a challenge. Web mining offers a solution by applying data mining techniques to extract user interests from web data, enabling personalized web experiences. This paper focuses on the application of web mining for personalizing web pages, particularly within the context of e-commerce, where effective recommendation systems are crucial for customer engagement and profitability. The authors identify a need for a flexible and efficient structure that can adapt to varying scales of online stores and user bases.

Methods

The paper proposes a mixed personalization system that moves beyond traditional reliance on access logs. Key aspects of their approach include:

  • User Identification: The system first identifies users, either through registration or by IP address for unregistered users.
  • Data Collection and Storage: Instead of solely relying on access logs, the proposed structure stores user-specific information (preferences, purchase history, viewed pages) directly in a database. This data is retained for the user's last five sessions to balance detail with storage efficiency.
  • Clustering: The system employs clustering techniques for ads, pages, and customers. This involves creating matrices (Products, Pages, Advertisements) that capture user interactions and preferences.
  • Similarity Measures: Formulas are provided to calculate the similarity between products, pages, and advertisements, as well as between customers. These measures are crucial for making relevant recommendations.
  • Decision Making: Based on the collected and analyzed data, the system makes decisions to personalize the user's experience, including displaying relevant advertisements and recommending products or pages.
  • Flexibility: The proposed structure is designed to be highly flexible, scalable from small online stores to large ones, with the primary difference being the specific algorithms used for grouping elements.

Datasets

The paper does not explicitly mention specific datasets used for training or evaluation. Instead, it describes the types of data that would be collected and utilized within their proposed structure, such as:

  • Access Logs: While the paper aims to move beyond sole reliance on these, they are mentioned as a potential source of information.
  • User Registration Data: Information provided by users during registration.
  • Database Records: Stored information about customers, products, advertisements, and pages.
  • Client-side Data: Cookies and potentially data from client-side applications.

Key Insights and Ideas

  • Beyond Access Logs: The paper highlights the limitations of relying solely on access logs for personalization, suggesting a more robust approach by directly storing and analyzing user interaction data in a database.
  • Dynamic Personalization: The proposed structure emphasizes dynamic personalization, where the web experience is tailored in real-time as the user interacts with the site.
  • Integrated Recommendation System: The structure inherently incorporates a recommendation system, aiming to suggest relevant items and pages to users.
  • Flexibility as a Core Feature: The emphasis on a flexible structure that can adapt to different scales of e-commerce operations is a significant design principle.
  • Balancing Accuracy and Speed: The authors acknowledge the trade-off between accuracy and speed in personalization and suggest that their structure aims to achieve a balance through various classification and similarity calculation methods.

Key Results

The paper describes the proposed structure and its components but does not present quantitative results from an implemented system or comparative studies. The "Evaluation" section discusses the potential benefits and how the structure aims to address existing problems, suggesting:

  • Improved Accuracy and Speedup: The structure is claimed to fix existing problems and provide both accuracy and speedup.
  • Adaptability: The system is designed to act "smarter" with more data, leading to better results for the customer.
  • Efficiency: The use of Boolean operators in classification formulas is expected to improve processing speed.
  • Scalability: The structure is deemed suitable for various web application languages and can be adapted for different sizes of e-commerce platforms.

Limitations and Future Work

  • No Quantitative Evaluation: The paper lacks empirical evaluation or quantitative results to validate the claimed improvements in accuracy and speed.
  • Data Storage for Large Sites: For very large websites (over 1000 users and 1000 pages), the authors suggest saving only the IDs of visited pages to manage storage.
  • Algorithm Dependency: The effectiveness of the structure is noted to depend on the specific algorithms used for classification and similarity calculations.
  • Future Work: The paper implies that further development and testing of the proposed structure would be beneficial. The authors suggest that the structure's efficiency can vary based on the underlying database and environment.

Summary updated

Abstract

During users visiting a Web Server Access, data is stored. This information can be used broadly. Using this information we can obtain users' preferences and use them to personalize Web pages. Web mining is a new discussion that has been proposed to manage the web pages. In fact Web mining is the application of data mining techniques to discover patterns of user interests is the Data Web. In this article we provide a structure for web mining.
BibTeX
@article{alinejadrokny2013proposing,
  author = {Hamid Alinejad‐Rokny and Mostafa Keikhay Farzaneh and Amir Goran Orimi and Mir Mohsen Pedram and Hojjat Ahangari Kiasari},
  title = {{Proposing a New Structure for Web Mining and Personalizing Web Pages}},
  journal = {Journal of Emerging Technologies in Web Intelligence},
  year = {2013},
  doi = {10.4304/jetwi.5.3.287-295},
}
Figure