Skip to main content

Web structure/content/usage mining; privacy

Short answer

Web content mining extracts knowledge from page contents; structure mining analyses hyperlinks using PageRank and HITS; usage mining discovers user patterns from logs. Privacy issues include collection without consent, inference, profiling and re-identification. Remedies include anonymisation, perturbation, differential privacy, consent policies and legal protection.

Explain Web Structure/Content/Usage mining. Explain the privacy aspects of Data Mining.

Web content mining

It extracts useful knowledge from the contents of web pages: text, images, audio, video and structured tables. It uses text mining, classification and clustering of documents, information extraction, summarisation and opinion mining. Example: classifying news pages into sports, politics, and so on; extracting product prices from shopping sites.

Web structure mining

It studies the hyperlink structure of the web, a graph with pages as nodes and links as edges, to find which pages are important and how pages are related.

  • PageRank: a page is important if important pages link to it.
  • HITS: finds hubs (pages linking to many good pages) and authorities (pages linked from many hubs).
  • Also community discovery, and focused crawling. Used in search engine ranking.

Web usage mining

It discovers user access patterns from web server logs, proxy logs, cookies and click-streams.

  1. Preprocessing: data cleaning, user identification, session and page-view identification.
  2. Pattern discovery: association rules, sequential patterns, clustering of users and classification.
  3. Pattern analysis: filtering and visualisation. Used for personalisation, recommendations, improving site design and caching, and fraud detection.
Web mining
  Content   -> what is on the page
  Structure -> how pages are linked
  Usage     -> how users behave

Privacy aspects of data mining

  • Personal information is collected and mined without informed consent (browsing history, purchases, location).
  • Combining data from many sources allows inference of sensitive facts and re-identification of "anonymous" data.
  • Profiling can lead to discrimination, and data can be misused or sold.
  • Web usage mining with cookies and logs tracks individuals across sites.
  • Remedies: privacy-preserving data mining (anonymisation, perturbation, differential privacy), consent and clear privacy policies, data minimisation, access control and encryption, and legal protection (e.g. the Privacy Act, 2075 of Nepal, GDPR).

✓ Answer checked by Claude Opus 5.5 (high effort)

Related questions

All Data Mining chapter 7 questions →