WEB MINING

[seminar class]

[attachment=9929]
INTELLIGENT WEB MINING
Improvising efficiency of web search engines
PRESENT SCENARIO
• INDEX BASED SEARCHING
• Page Ranking Algorithm.
• PROBLEMS
• DIFFICULT FOR INEXPERIENCED USERS.(You don’t get what you want!)
• POLYSEMY PROBLEM DUE TO INDEX SEARCH.
• POSSIBLE FLAW IN PAGE RANKING ALGORITHM.
REAL TIME EXAMPLE
• SEARCH FOR “BUSH” IN WWW.GOOGLE.COM
• 266,000,000 RESULTS!!!
• FIRST TEN PAGES ONLY HAS PRESIDENT BUSH.
• IS THERE ONLY PRESIDENT BUSH IN THIS WORLD?
HOW SEARCH ENGINES WORK?
• BY ANALYSING WEB PAGE STRUCTURE, USING DOM TREE STRUCTURE.
OUR PROPOSALS
TO OVER COME THE PRESENT PROBLEMS :
• CONTEXT BASED SEARCH
• DUAL ROLE TREE STRUCTURE
• TAGGING SIMILAR WORDS TOGETHER
• CONTEXT BASED SEARCH
• IDENTIFIES CONTEXTS IN WEB PAGES THROUGH AUTOMATED KEYWORD IDENTIFICATION.
• CONTEXT WORDS BECOME NODES OF CONTEXT BASED TREE.
• NODES ARE ORDERED BASED ON SIMILARITY WITH KEYED IN WORD.
• SEARCH ENGINE SEARCHES CONTEXT TREE.
• DISPLAY.
• DUAL ROLE BASED TREE
DOM TREE STRUCTURE
CONTEXT BASED
TREE STRUCTURE
BUT,HOW TO CREATE CONTEXTS?
ANT ANALOGY
 ANT IDENTIFIES SMELL OF FOOD.
HERE SMELL IS ATTRIBUTE.
SIMILARLY IDENTIFY ATTRIBUTES OF DATA.
SEARCH FOR THEM.
 ANT LOOKS IN LIKELY PLACES.
SIMILARLY, SEARCH FOR
LIKELY CLUSTERS

USING:
CORRELATION
ANALYSIS.
• PROTOTYPE
Search for : BUSH
• PRESIDENT,SHRUBS,TRIBES(BUSHMEN) could be possible nodes of context tree.
• PRIORITY WOULD BE GIVEN FOR EVERY NODE.
• CHANCES LIKELY THAT USER IS NOT DISAPPOINTED.
BENEFITS
• EFFICIENT
• UN-NECESSARY INFORMATION WILL BE ABSENT.
• IMPROVISES PAGE RANKING ALGORITHM.