Data Mining Core Concept & Introduction.

Data Mining Core Concept & Introduction.

• 1.
  Data Mining By Engr. AhsanShah
• 2.
  Example Explanation 🛒Market BasketRule: “If a customer buys bread, they also buy butter.” Association pattern 📈Sales increase during holidays. Trend pattern 👩‍⚕️Certain symptoms often appear together in patients. Correlation pattern 💳Unusually high transaction = possible fraud. Anomaly pattern WhatisData Mining? Data Mining is the process of discovering useful patterns, relationships, and insights from large sets of data using statistical,mathematical,andcomputationaltechniques. In Data Mining, a pattern means a useful, meaningful, and valid relationshiporstructurefoundindata. These patterns help understand behavior, predict outcomes, andmakebusinessdecisions.
• 3.
  Transaction ID ItemsBought T1 Milk, Bread, Butter T2 Milk, Bread T3 Bread, Butter T4 Milk, Bread, Butter, Eggs AssociationPatterns (DependencyRules) Showsrelationshipsbetweenitemsorvariables. Commoninmarketbasketanalysis. If{Milk,Bread}→{Butter}(Support=30%,Confidence=80%) 30%oftransactionsincludemilk,bread,andbutter; 80%ofcustomerswhobuymilkandbreadalsobuybutter. TechniqueUsed AprioriAlgorithm FP-GrowthAlgorithm WhatistheAprioriAlgorithm? TheAprioriAlgorithmisaclassicdataminingalgorithmused tofindfrequentitemsetsandgenerateassociationrulesfrom largetransactionaldatabases. Findpatternslike “IfacustomerbuysMilkandBread,theyalsobuy Butter.” Before understandinghow Aprioriworks,you mustknowthese threeimportant terms: 1.Itemset Asetofitemsboughttogetherinatransaction. Example:{Milk,Bread} 2.Support Measureshowfrequentlyanitemset appearsinthedataset. Support(A→B)=Totaltransactions / Transactionscontaining(A∪B)​ If2outof4transactionscontain{Milk,Bread}, Support=2/4=0.5=50%
• 4.
  Whatdoestheformulamean? ItisusedtomeasuretheconfidenceoftheruleA⇒B,whichasks: "HowlikelyisBtobetruewhenAistrue?" Support(A): TheproportionoftransactionsinthedatasetthatcontainitemsetA. Support(A∪B): TheproportionoftransactionsthatcontainbothAandBtogether. Confidence(A→B): ThelikelihoodthatatransactioncontainingAalsocontainsB. It'scalculatedby: Confidence(A→B) = TransactionswithbothAandB​ Transactionswith A Supposeyouhave100transactionsfromastore: 40containmilk(A) 30containbothmilkandbread(A∪B) Support(A)=40/100=0.40 Support(A∪B)=30/100=0.30 Confidence(A→B)=0.30/0.40=0.75 So,75%ofthecustomerswhoboughtmilkalsoboughtbread. Theconfidenceformula helpsquantifythe strengthofarulein associationmining.A higherconfidence meansastrongerrule— i.e.,ifAhappens,Bis morelikelytohappen. 3.Confidence
• 5.
  Candidate Support (%) {Milk,Bread} 75% {Milk, Butter} 50% {Bread, Butter} 75% Candidate Support (%) {Milk, Bread, Butter} 50% Rule Support Confidence Lift Milk → Bread 75% 100% 1 Bread → Butter 75% 75% 1 Milk, Bread → Butter 50% 66% >1 4.Lift ✅IfLift>1→positiverelationship ✅IfLift=1→norelationship ✅IfLift<1→negativerelationship StepsintheAprioriAlgorithm Step1:SetMinimumSupportandConfidence Decidethresholds(e.g.,support≥50%,confidence≥60%). Step2:FindAllFrequent1-Itemsets Countsupportforeachsingleitem. Keeponlyitemsmeetingtheminimumsupportthreshold. Step3:GenerateCandidate2-Itemsets Step4:GenerateCandidate3-Itemsets Combinefrequent2-itemsets. Step5:GenerateAssociationRules
• 6.
  Step Description ExampleOutput 1 Choose min support/confidence Support ≥ 50%, Confidence ≥ 60% 2 Generate 1-itemsets {Milk}, {Bread}, {Butter} 3 Generate 2-itemsets {Milk, Bread}, {Bread, Butter} 4 Generate 3-itemsets {Milk, Bread, Butter} 5 Generate rules Milk & Bread → Butter
• 7.
  Domain Application 🛍️Retail Market BasketAnalysis (find products bought together) 🏦Banking Detecting services used together 🌐Web Usage Mining Finding patterns in website navigation 🧬Healthcare Identifying symptom-disease relationships 📧Marketing Cross-selling and recommendation systems Real-LifeApplications ofApriori
• 8.
  Advantages andLimitations Advantages Limitations Simpleandeasytoimplement Workswellforsmalltomediumdata Providesclearrules Highcomputationcostforlargedatasets Generatestoomanycandidatesets Doesn’thandlenumericorcontinuousdataeasily
• 9.
  Transaction ID ItemsBought T1 Milk, Bread, Butter T2 Bread, Butter T3 Milk, Bread T4 Milk, Bread, Butter FP-Growth Algorithm? Insteadofgeneratingcandidateitemsetsonebyone(likeApriori), FP-GrowthusesacompactdatastructurecalledtheFP-Tree (FrequentPatternTree). Point01 Scanthetransactiondatabaseoncetofindfrequentitems. Sortitemsineachtransactionbytheirfrequency. Buildatreestructurethatstoresitemsandtheiroccurrence counts. Step1:BuildtheFP-Tree FP-Growth (Frequent Pattern Growth) is a data mining algorithm used to find frequent itemsets in large datasets just like the Apriori algorithm, but faster and more efficient. Step2:ExtractFrequentItemsets Startingfromthebottomofthetree,recursivelyfindprefix paths(patterns). GenerateconditionalFP-Treesforeachitem. Combinethemtoformfrequentitemsets.
• 10.
  Limitation Description 🧩Complex TreeStructure Can be hard to understand and 🧮Memory Usage May grow large for sparse data (many 🧠Not Easy for Dynamic Data Tree must be rebuilt if data Root ├──Bread(4) │ ├──Milk(3) │ └──Butter(3) Then,frequentpatternsareextractedsuchas: {Bread} {Bread,Milk} {Bread,Butter} {Bread,Milk,Butter}
• 11.
  Step2:KDDProcess (KnowledgeDiscoveryin Database) DataCleaning–Removenoiseandmissingvalues DataIntegration–Combinedatafrommultiplesources DataSelection–Chooserelevantdata DataTransformation–Convertintosuitableformat DataMining–Applyalgorithmsandfindpatterns PatternEvaluation–Identifymeaningfulresults KnowledgePresentation–Visualizeresultsforusers Beforemining,rawdatamustbecleaned,organized,andtransformed. ThisprocessiscalledDataPreprocessing. Step3:PreparingtheData (DataPreprocessing) StepsinDataPreparation: 1.DataCleaning Removeduplicate,missing,orinconsistentdata. Example:Replacingblankvalueswiththeaverageofthecolumn. 2.DataIntegration Combinedatafrommultiplesources(e.g.,sales+customer+locationdata). Helpscreateaunifieddataset. 3.DataSelection Chooseonlyrelevantattributesformining. Example:Forpredictingsales,choose“price,”“discount,”“region”butnot“employeeage 4.DataTransformation Convertdataintoasuitableformatforanalysis. Includes: Normalization:Scalevaluesintoafixedrange(e.g.,0–1) Aggregation:Summarizedata(e.g.,weekly→monthlysales) Encoding:Converttextdataintonumbersforalgorithms. Tomakethedatasetaccurate,consistent,and readyforminingalgorithms. Poordataquality=poorresults,nomatterhow goodthealgorithmis.