Discovering potential user browsing behaviors using custom-built apriori algorithm

Most of the organizations put information on the web because they want it to be seen by the world. Their goal is to have visitors come to the site, feel comfortable and stay a while and try to know completely about the running organization. As educational system increasingly requires data mining, the opportunity arises to mine the resulting large amounts of student information for hidden useful information (patterns like rule, clustering, and classification, etc). The education domain offers ground for many interesting and challenging data mining applications like astronomy, chemistry, engineering, climate studies, geology, oceanography, ecology, physics, biology, health sciences and computer science. Collecting the interesting patterns using the required interestingness measures, which help us in discovering the sophisticated patterns that are ultimately used for developing the site. We study the application of data mining to educational log data collected from Guru Nanak Institute of Technology, Ibrahimpatnam, India. We have proposed a custom-built apriori algorithm to find the effective pattern analysis. Finally, analyzing web logs for usage and access trends can not only provide important information to web site developers and administrators, but also help in creating adaptive web sites.

💡 Research Summary

This research paper, titled “Discovering potential user browsing behaviors using custom-built apriori algorithm,” presents a study on applying web usage mining techniques to analyze server log data from an educational institution’s website. The primary objective is to uncover hidden user navigation patterns that can inform website improvements and personalization.

The authors focus on the Guru Nanak Institute of Technology’s web logs. They argue that as educational systems increasingly rely on online platforms, mining the large volumes of resulting user data becomes crucial for extracting useful patterns like association rules, which can reveal how users interact with the site.

The paper begins by contextualizing web mining within its three sub-areas: content, structure, and usage mining. It positions its work within web usage mining, which involves analyzing secondary data like IP addresses, access times, and URLs from server logs to understand user behavior. The introduction highlights potential applications such as enhancing website organization, improving marketing effectiveness, and creating adaptive websites.

A significant portion of the paper reviews related work, citing numerous previous studies on web log mining, improved algorithms (like AprioriAll variants), and frameworks for personalization. This establishes the foundation upon which the authors build their contribution.

The core technical contribution is the proposal of a “custom-built apriori algorithm.” The standard Apriori algorithm is a classic method for association rule mining. The authors modify it to better suit the characteristics of web log data. Their algorithm processes distinct IP addresses and URLs, filtering for “successful” visits based on HTTP status codes. It iteratively performs join and prune steps to generate frequent itemsets (e.g., combinations of an IP and a URL) that exceed a predefined threshold of hits (“support”). From these frequent itemsets, association rules (like “if IP address A, then URL B”) are generated.

The authors demonstrate their methodology using real log data. They present results in three categories:

1. General Statistics: Provides an overview, including total hits, number of visitors, successful visits, and various error reports.
2. Access Statistics: Offers a more detailed breakdown of successful and unsuccessful hits based on specific IP addresses and URLs.
3. Co-relations (Rules): Displays the association rules discovered by their custom algorithm. These rules depict relationships between IP addresses, URLs, and file paths (e.g., ipadd -> url, url -> path, and the more complex ipadd -> url -> path). These rules effectively map common user navigation pathways through the website.

A key part of the results is a comparative table that outlines the procedural steps of the standard Apriori algorithm alongside their custom version. This comparison aims to show how their algorithm is adapted for the specific task, though a rigorous quantitative evaluation of performance gains (e.g., in speed or scalability) is not deeply explored.

In conclusion, the paper asserts that the proposed custom-built Apriori algorithm is effective for analyzing educational web log files. It successfully discovers various co-relations (rules) among user data items within a reasonable execution time. The generated patterns provide actionable insights for web developers and administrators to optimize site structure, enhance user experience, and move towards creating more adaptive web environments. The study underscores the practical value of data mining in the educational domain for web resource management.