3618

3416

كارشناسي ارشد

كامپيوتر

اصفهان: دانشگاه صنعتي اصفهان، دانشكده برق و كامپيوتر

1385

نه، 122، [II] ص.: مصور، جدول، نمودار

ص. ع. به فارسي و انگليسي

محمدحسين سرائي

مازيار پالهنگ

فارسي به انگليسي و انگليسي به فارسي

7/5/86

محمد داورپناه، جواد عسكري

مهندسي برق و كامپيوتر

ID3416

به فارسي و انگليسي: قابل رويت در نسخه ديجيتال

AbstractWith the recent developments in computer storage and information technology massiveamounts of data have been stored and it is very difficult to process them without automaticdata analysis methods Large amount of knowledge exist within these datasets but hiddenfrom the users Data mining is the process of extraction of unknown knowledge from largeamounts of data Knowledge is a concept beyond data and information Knowledge isfinding patterns and hidden trends among data and information Different types of rulescan be discovered by the process of data mining including association rules classificationand generalization rules In this thesis we present mining association rules which aims atdiscovering the patterns of co occurrences of the attributes in the database In this approachof data mining potentially interesting relationships and associations between data arerecognized and presented to the user An example of such a rule is might be that 98 ofcustomers that purchase tires and automotive accessories also get automotive servicesdone Finding all such rules is valuable foe cross marketing and attached mailingapplications Other applications include catalog design store layout and customersegmentation based on buying patterns It has been discovered recently that time dependent information is important in datamining So temporal patterns or rules should be discovered from temporal data since it canprovide accurate information about an evolving business domain rather than a static onethat conventional data mining is dealing with Time is considered in discussed miningmethods in this thesis There are many time aspects where can associate with rules One ofthese aspects is time interval A time interval associate with each association rule showingwhen that rule is valid Thus knowing time interval of a pattern usefulness of extractedknowledge is increased Another time aspect where can associate with rules is time period A series of repeated occurrences of a certain type of event at regular intervals is describedas a periodic event In this thesis both time interval and time period factors is discussed inmining temporal association rules We extend one of existing teqniques in order to create a flexible schema for expressingvalidation time of the rules This system uses calendar based times for discoveringassociation rules Generated rules could have exact and usefull concept in real applications We extended this schema to present time periods in more flexible ways Time periods inproposed schema have more flexibility and can express more complex time periods inaddition to time periods of the existing approach Furthermore proposed schema is able touse time intervals too Thus the proposed schema led to more precision of each rulevalidation time resulting in more usefulness of rules Furthermore implementation resultshows that the proposed approach generated more and potentially more interestingassociation rules compared to the existing approaches

محمدحسين سرائي

مازيار پالهنگ

محمد داورپناه، جواد عسكري