Presentation on theme: "Intelligence Through Learning from Data"— Presentation transcript:
1 Intelligence Through Learning from Data Monash UniversitySemester 1, March 2006
2 Lecture OutlineMachine Learning – Yet another form of intelligent softwareLearning for DataData Mining – A real world application of learning from dataData Mining ConceptsData Mining TechniquesData Mining Applications
3 Lecture Objectives By the end of this lecture, you should: Understand the relationship between machine learning and data miningKnow the principles of learning from data and the various techniques for learning from dataUnderstand the real world applications of learning from dataBe able to distinguish between this form of intelligence in software systems and other strategies such as software agents, context-awareness, expert systems and knowledge representation/deductive approachesWhy association rules - business needs to identify the relationshipsMachine learning – effective data mining requires learning.
5 Machine LearningMachine Learning is an area of Artificial Intelligence.It is concerned with programs that learnData Mining uses machine learning for prediction and classificationFeedback on the correctness of a prediction combined with examples and domain knowledge allow the program to learn.Machine Learning is also used in speech recognition, robot training, classification of astronomical structures and game playing.
6 Machine Learning“A general law can never be verified by a finite number of observations. It can, however, be falsified by only one observation.” Karl PopperThe patterns that machine learning algorithms find can never be definitive theoriesAny results discovered must to be tested for statistical relevance
7 The Empirical CycleAnalysisTheoryObservationPrediction
8 Concept Learning - 1 Example: the concept of a wombat a learning algorithm could consider many animals and be advised in each case whether it is a wombat or not. From this a definition would be deduced.The definition iscomplete if it recognizes all instances of a concept ( in this case a wombat).consistent if it does not classify any negative examples as falling under the concept.
9 Concept Learning - 2An incomplete definition is too narrow and would not recognize some wombats.An inconsistent definition is too broad and would classify some non-wombats as wombats.A bad definition could be both inconsistent and incomplete.
10 Hypothesis Characteristics - 1 Classification Accuracy1 in a million wrong is better than 1 in 10 wrong.TransparencyA person is able understand the hypothesis generated. It is then much easier to take action
11 Hypothesis Characteristics - 2 Statistical SignificanceThe hypothesis must perform better than the naïve prediction. (Imagine if 80% of animals considered are wombats and the theory is that all animals are wombats then the theory is right 80% of the time! But nothing has been learnt.)Information ContentWe look for a rich hypothesis. The more information contained (while still being transparent) the more understanding is gained and the easier it is to formulate an action plan.
12 Complexity of Search Space Machine learning can be considered as a search problem. We wish to find the correct hypothesis from among many.If there are only a few hypotheses we could try them all but if there are an infinite number we need a better strategy.If we have a measure of the quality of the hypothesis we can use that measure to select potential good hypotheses and based on the selection try to improve the theories (hill-climbing search)Consider the metaphor of the kangaroo in the mist.This demonstrates that it is important to know the complexity of the search space. Also that some pattern recognition patterns are almost impossible to solve.
13 Learning as a Compression We have learnt something if we have an algorithm that creates a description of the data that is shorter than the original data setA knowledge representation is required that is incrementally compressible and an algorithm that can achieve that incremental compressionThe file-in could be a relation table and the file-out a prediction or a suggested clusteringFile-outAlgorithmFile-in
15 Introduction Motivation: Why data mining? What is data mining? Data Mining: On what kind of data?Data mining functionalityAre all the patterns interesting?Classification of data mining systemsLink to Data Warehousing
16 Motivation: “Necessity is the Mother of Invention” Data explosion problemAutomated data collection tools and mature database technology lead to tremendous amounts of data stored in databases, data warehouses and other information repositoriesWe are drowning in data, but starving for knowledge!Solution: Data warehousing and data miningData warehousing and on-line analytical processingExtraction of interesting knowledge (rules, regularities, patterns, constraints) from data in large databases
17 Evolution of Database Technology Data collection, database creation, IMS and network DBMS1970s:Relational data model, relational DBMS implementation1980s:RDBMS, advanced data models (extended-relational, OO, deductive, etc.) and application-oriented DBMS (spatial, scientific, engineering, etc.)1990s—2000s:Data mining and data warehousing, multimedia databases, and Web databases
18 What Is Data Mining?Data mining (knowledge discovery in databases - KDD):Extraction of interesting (non-trivial, implicit, previously unknown and potentially useful) information or patterns from data in large databasesAlternative names and their “inside stories”:Data mining: a misnomer?Knowledge discovery(mining) in databases (KDD), knowledge extraction, data/pattern analysis, data archeology, data dredging, information harvesting, business intelligence, etc.What is not data mining?(Deductive) query processing.Expert systems or small ML/statistical programs
19 Rules 1% support If test A then test B Episodes DatabaseGP DatabaseData PreparationMergeAssociation DiscoveryDatabase SegmentationRules 1% support If test A then test Bwill occur in 62%of casesSegment 1 Segment 297 GPs GPsScore = 1.8 Score = 2.7
20 Why Data Mining? — Potential Applications Database analysis and decision supportMarket analysis and managementtarget marketing, customer relation management, market basket analysis, cross selling, market segmentationRisk analysis and managementForecasting, customer retention, improved underwriting, quality control, competitive analysisFraud detection and managementOther ApplicationsText mining (news group, , documents) and Web analysis.Intelligent query answering
21 Market Analysis and Management (1) Where are the data sources for analysis?Credit card transactions, loyalty cards, discount coupons, customer complaint calls, plus (public) lifestyle studiesTarget marketingFind clusters of “model” customers who share the same characteristics: interest, income level, spending habits, etc.Determine customer purchasing patterns over timeConversion of single to a joint bank account: marriage, etc.Cross-market analysisAssociations/co-relations between product salesPrediction based on the association information
22 Market Analysis and Management (2) Customer profilingdata mining can tell you what types of customers buy what products (clustering or classification)Identifying customer requirementsidentifying the best products for different customersuse prediction to find what factors will attract new customersProvides summary informationvarious multidimensional summary reportsstatistical summary information (data central tendency and variation)
23 Corporate Analysis and Risk Management Finance planning and asset evaluationcash flow analysis and predictioncontingent claim analysis to evaluate assetscross-sectional and time series analysis (financial-ratio, trend analysis, etc.)Resource planning:summarize and compare the resources and spendingCompetition:monitor competitors and market directionsgroup customers into classes and a class-based pricing procedureset pricing strategy in a highly competitive market
24 Fraud Detection and Management (1) Applicationswidely used in health care, retail, credit card services, telecommunications (phone card fraud), etc.Approachuse historical data to build models of fraudulent behavior and use data mining to help identify similar instancesExamplesauto insurance: detect a group of people who stage accidents to collect on insurancemoney laundering: detect suspicious money transactions (US Treasury's Financial Crimes Enforcement Network)medical insurance: detect professional patients and ring of doctors and ring of references
25 Fraud Detection and Management (2) Detecting inappropriate medical treatmentHealth Insurance Commission identifies that in many cases blanket screening tests might have been requested (can save $$).Detecting telephone fraudTelephone call model: destination of the call, duration, time of day or week. Analyze patterns that deviate from an expected norm.British Telecom identified discrete groups of callers with frequent intra-group calls, especially mobile phones, and broke a multimillion dollar fraud.RetailAnalysts estimate that 38% of retail shrink is due to dishonest employees.
26 Other Applications Sports IBM Advanced Scout analyzed NBA game statistics (shots blocked, assists, and fouls) to gain competitive advantage for New York Knicks and Miami HeatAstronomyJPL and the Palomar Observatory discovered 22 quasars with the help of data miningInternet Web Surf-AidIBM Surf-Aid applies data mining algorithms to Web access logs for market-related pages to discover customer preference and behavior pages, analyzing effectiveness of Web marketing, improving Web site organization, etc.
27 Data Mining: A KDD Process KnowledgeData mining: the core of knowledge discovery process.Pattern EvaluationData MiningTask-relevant DataData WarehouseSelectionData CleaningData IntegrationDatabases
28 The Process of Knowledge Discovery ReportingDataCleaning & EnrichmentCodingData miningselection- clustering-domain consistency- segmentation-de-duplication- prediction-disambiguationInformationRequirementActionFeedbackOperational dataExternal dataThe Knowledge Discovery in Databases (KDD) process (Adriens/Zantinge)
29 Steps of a KDD Process Learning the application domain: relevant prior knowledge and goals of applicationCreating a target data set: data selectionData cleaning and preprocessing: (may take 60% of effort!)Data reduction and transformation:Find useful features, dimensionality/variable reduction, invariant representation.Choosing functions of data miningsummarization, classification, regression, association, clustering.Choosing the mining algorithm(s)Data mining: search for patterns of interestPattern evaluation and knowledge presentationvisualization, transformation, removing redundant patterns, etc.Use of discovered knowledge
30 Data Mining and Business Intelligence Increasing potentialto supportbusiness decisionsMakingDecisionsEnd UserData PresentationBusinessAnalystVisualization TechniquesData MiningInformation DiscoveryDataAnalystData ExplorationStatistical Analysis, Querying and ReportingData Warehouses / Data MartsOLAP, MDAData SourcesDBAPaper, Files, Information Providers, Database Systems, OLTP
31 Architecture of a Typical Data Mining System Graphical user interfacePattern evaluationData mining engineKnowledge-baseDatabase or data warehouse serverFilteringData cleaning & data integrationDataWarehouseDatabases
32 Data Mining: On What Kind of Data? Relational databasesData warehousesTransactional databasesAdvanced DB and information repositoriesObject-oriented and object-relational databasesSpatial databasesTime-series data and temporal dataText databases and multimedia databasesHeterogeneous and legacy databasesWWW
33 Data Mining Techniques Various taxonomies exist. Berry & Linoff define 6 tasksClassificationEstimationPredictionClusteringDescriptionAffinity GroupingCabena et al. define 4 operations(i.e. tasks)Predictive ModelingDatabase SegmentationLink AnalysisDeviation DetectionBriefly we will look at the techniques and we will talk about one of them, how is done in detail in subsequent lectures. However it is worth while to know the taxonomy.Taxonomy can be done based on the principles used in mining properties of data or pre-conceived strategy of data mining.
34 ClassificationClassification involves considering the features of some object then assigning it it to some pre-defined class, for example:Spotting fraudulent insurance claimsWhich phone numbers are fax numbersWhich customers are high-valueBasically partition the data into groups so that the elements of each group has some common properties.Usually the properties are based on the features or derivative of the attributes of the objectsFor example, in the case of fraudulent data, may the withdrawn rate (time interval between the withdrawal) and/or the amount may constitute the feature for classification.
35 EstimationEstimation deals with numerically valued outcomes rather than discrete categories as occurs in classification.Estimating the number of children in a familyEstimating family incomeEstimation – primarily applied to numerical data rather than categorical dataEstimation is determine a value for an unknown output attribute.Most of the supervised learning algorithms can perform classification and estimationWe say a method is supervised learning when we know the dependent variable to guide the data mining (learning) process.For example, the classification is supervised learning because we know the attributes that we will use in defining the classes.On the contrary unsupervised leaning we are without a dependent variable.
36 PredictionEssentially the same as classification and estimation but involves future behaviourHistorical data is used to build a model explaining behaviour (outputs) for known inputsThe model developed is then applied to current inputs to predict future outputsPredict which customers will respond to a promotionClassifying loan applicationsDifficult to differentiate between classification and estimation.Purpose of prediction is to determine future outcome rather than current behaviour.The output variable of predictive model can be numeric or categorical.
37 ClusteringClustering is also sometimes referred to as segmentation (though this has other meanings in other fields)In clustering there are no pre-defined classes. Self-similarity is used to group records. The user must attach meaning to the clusters formedClustering often precedes some other data mining task, for example:once customers are separated into clusters, a promotion might be carried out based on market basket analysis of the resulting clusterClustering can be supervised or unsupervised.In the case of unsupervised clustering – is to discover concept structures in data.
38 DescriptionA good description of data can provide understanding of behaviourThe description of the behaviour can suggest an explanation for it as wellStatistical measures can be useful in describing data, as can techniques that generate rules
39 Deviation DetectionRecords whose attributes deviate from the norm by significant amounts are also called outliersApplication areas include:fraud detectionquality controltracing defects.Visualization techniques and statistical techniques are useful in finding outliersA cluster which contains only a few records may in fact represent outliersOutliers – quite useful when you want to identify unusual or non-standard data.Visualisation techniques are quite useful here as the outlier data will stand out from other data.Or clustering where the size of the cluster may give some clue.
40 Affinity GroupingAffinity grouping is also referred to as Market Basket AnalysisA common example is the discovery of which items are frequently sold together at a supermarket. If this is known, decisions can be made about:arranging items on shelveswhich items should be promoted togetherwhich items should not simultaneously be discountedAlso referred as relationship analysis – but popularly known as market basket analysis because of its use in retail business applications.Once you identify the relationships, various strategies can be used to exploit the relationship to the benefit of the business.
41 Association Rule Mining ConfidenceRule BodyWhen a customer buys a shirt, in 70% of cases,he or she will also buy a tie!We find this happens in 13.5% of all purchases.How do you specify the relationship – by means of rules (also called association rules)Association rule has a number of components in simple case there will be a LHS and RHS and they are related.Formally expressed as IF … THEN … rule.Rule HeadSupport
42 Association Rule Mining Some rules are useful: Unknown, unexpected and indicative of some action to take.Some rules are trivial: Known by anyone familiar with the business.Some rules are inexplicable: Seem to have no explanation and do not suggest a course of action. “The key to success in business is to know something that nobody else knows” Aristotle Onassis
43 Co-Occurrence Table Customer Items 1 orange juice (OJ), cola 2 milk, orange juice, window cleaner3 orange juice, detergent4 orange juice, detergent, cola5 window cleaner, colaOJ Cleaner Milk Cola DetergentOJCleanerMilkColaDetergentLet me explain by an example before we formalise the association rule.
44 From the Co-Occurrence Table We can say that people who buys Orange Juice also will buy Cola ( or detergent) orange juice colaThis association rule is satisfied by 2 out of 5 customers ( 1 and 4) hence support is 2/5 = 40%However, there are three customers (1,3 and 4) have purchased orange juice and hence the confidence of the above rule is only 2/3 = 66.67%Question: Are support and confidence measures good enough?The rule has one item (or attribute) on the left hand side and the right hand side. How do you find rules which has more than one items on the left hand side (multi-attribute rule)What it means when the support is 100% - an obvious rule of no significance to the business.Some support but low confidence!!Low support but high confidence?
45 Support and Confidence Percentage of transactions from a transaction database that the given rule satisfies.This can be taken as the probability P(X Y) where X Y indicates that a transaction contains both X and Y, that is union of item sets X and Y.Confidence:Which assess the degree of certainty of the detected association.This can be taken as the conditional probability P(Y|X), that is, the probability that a transaction containing X also contains Y.More formallySupport (X Y ) = P (X Y)Confidence (X Y) = P (Y|X)
46 What is a Rule? If condition then result Note: If nappies and Thursday then beeris usually better than (in the sense that it is more actionable)If Thursday then nappies and beer because it has just one item in the resultIf a 3 way combination is the most common, then consider rules with just 1 item in the result, e.g.If A and B, then CIf A and C, then BSo far we talk rule of one LHS and one RHS.It does not matter how many attributes we have on RHS, they can be broken into canonical form.What about more than attribute on LHS?
47 Data Mining Functionalities (2) Classification and PredictionFinding models (functions) that describe and distinguish classes or concepts for future predictionE.g., classify countries based on climate, or classify cars based on gas mileagePresentation: decision-tree, classification rule, neural networkPrediction: Predict some unknown or missing numerical valuesCluster analysisClass label is unknown: Group data to form new classes, e.g., cluster houses to find distribution patternsClustering based on the principle: maximizing the intra-class similarity and minimizing the interclass similarity
48 Data Mining Functionalities (3) Outlier analysisOutlier: a data object that does not comply with the general behavior of the dataIt can be considered as noise or exception but is quite useful in fraud detection, rare events analysisTrend and evolution analysisTrend and deviation: regression analysisSequential pattern mining, periodicity analysisSimilarity-based analysisOther pattern-directed or statistical analyses
49 Are All the “Discovered” Patterns Interesting? A data mining system/query may generate thousands of patterns, not all of them are interesting.Suggested approach: Human-centered, query-based, focused miningInterestingness measures: A pattern is interesting if it is easily understood by humans, valid on new or test data with some degree of certainty, potentially useful, novel, or validates some hypothesis that a user seeks to confirmObjective vs. subjective interestingness measures:Objective: based on statistics and structures of patterns, e.g., support, confidence, etc.Subjective: based on user’s belief in the data, e.g., unexpectedness, novelty, actionability, etc.
50 Can We Find All and Only Interesting Patterns? Find all the interesting patterns: CompletenessCan a data mining system find all the interesting patterns?Association vs. classification vs. clusteringSearch for only interesting patterns: OptimizationCan a data mining system find only the interesting patterns?ApproachesFirst general all the patterns and then filter out the uninteresting ones.Generate only the interesting patterns—mining query optimization
51 Data Mining: Confluence of Multiple Disciplines DatabaseTechnologyStatisticsData MiningMachineLearningVisualizationInformationScienceOtherDisciplines
52 Data Mining: Classification Schemes General functionalityDescriptive data miningPredictive data miningDifferent views, different classificationsKinds of databases to be minedKinds of knowledge to be discoveredKinds of techniques utilizedKinds of applications adapted
53 A Multi-Dimensional View of Data Mining Classification Databases to be minedRelational, transactional, object-oriented, object-relational, active, spatial, time-series, text, multi-media, heterogeneous, legacy, WWW, etc.Knowledge to be minedCharacterization, discrimination, association, classification, clustering, trend, deviation and outlier analysis, etc.Multiple/integrated functions and mining at multiple levelsTechniques utilizedDatabase-oriented, data warehouse (OLAP), machine learning, statistics, visualization, neural network, etc.Applications adaptedRetail, telecommunication, banking, fraud analysis, DNA mining, stock market analysis, Web mining, Weblog analysis, etc.
54 Data Mining and the Data Warehouse Organizations realized that they had large amounts of data stored (especially of transactions) but it was not easily accessibleThe data warehouse provides a convenient data source for data mining. Some data cleaning has usually occurred. It exists independently of the operational systemsData is retrieved rather than updatedIndexed for efficient retrievalData will often cover 5 to 10 yearsA data warehouse is not a pre-requisite for data mining
55 Data Mining and OLAP Online Analytic Processing (OLAP) Tools that allow a powerful and efficient representation of the dataMakes use of a representation known as a cubeA cube can be sliced and dicedOLAP provide reporting with aggregation and summary information but does not reveal patterns, which is the purpose of data mining
56 Major Issues in Data Mining (1) Mining methodology and user interactionMining different kinds of knowledge in databasesInteractive mining of knowledge at multiple levels of abstractionIncorporation of background knowledgeData mining query languages and ad-hoc data miningExpression and visualization of data mining resultsHandling noise and incomplete dataPattern evaluation: the interestingness problemPerformance and scalabilityEfficiency and scalability of data mining algorithmsParallel, distributed and incremental mining methods
57 Major Issues in Data Mining (2) Issues relating to the diversity of data typesHandling relational and complex types of dataMining information from heterogeneous databases and global information systems (WWW)Issues related to applications and social impactsApplication of discovered knowledgeDomain-specific data mining toolsIntelligent query answeringProcess control and decision makingIntegration of the discovered knowledge with existing knowledge: A knowledge fusion problemProtection of data security, integrity, and privacy
58 SummaryData mining: discovering interesting patterns from large amounts of dataA natural evolution of database technology, in great demand, with wide applicationsA KDD process includes data cleaning, data integration, data selection, transformation, data mining, pattern evaluation, and knowledge presentationMining can be performed in a variety of information repositoriesData mining functionalities: characterization, discrimination, association, classification, clustering, outlier and trend analysis, etc.Classification of data mining systemsMajor issues in data mining
59 A Brief History of Data Mining Society 1989 IJCAI Workshop on Knowledge Discovery in Databases (Piatetsky-Shapiro)Knowledge Discovery in Databases (G. Piatetsky-Shapiro and W. Frawley, 1991)Workshops on Knowledge Discovery in DatabasesAdvances in Knowledge Discovery and Data Mining (U. Fayyad, G. Piatetsky-Shapiro, P. Smyth, and R. Uthurusamy, 1996)International Conferences on Knowledge Discovery in Databases and Data Mining (KDD’95-98)Journal of Data Mining and Knowledge Discovery (1997)1998 ACM SIGKDD, SIGKDD’ conferences, and SIGKDD ExplorationsMore conferences on data miningPAKDD, PKDD, SIAM-Data Mining, (IEEE) ICDM, etc.
60 Where to Find References? Data mining and KDD (SIGKDD member CDROM):Conference proceedings: KDD, and others, such as PKDD, PAKDD, etc.Journal: Data Mining and Knowledge DiscoveryDatabase field (SIGMOD member CD ROM):Conference proceedings: ACM-SIGMOD, ACM-PODS, VLDB, ICDE, EDBT, DASFAAJournals: ACM-TODS, J. ACM, IEEE-TKDE, JIIS, etc.AI and Machine Learning:Conference proceedings: Machine learning, AAAI, IJCAI, etc.Journals: Machine Learning, Artificial Intelligence, etc.Statistics:Conference proceedings: Joint Stat. Meeting, etc.Journals: Annals of statistics, etc.Visualization:Conference proceedings: CHI, etc.Journals: IEEE Trans. visualization and computer graphics, etc.
61 ReferencesU. M. Fayyad, G. Piatetsky-Shapiro, P. Smyth, and R. Uthurusamy. Advances in Knowledge Discovery and Data Mining. AAAI/MIT Press, 1996.J. Han and M. Kamber. Data Mining: Concepts and Techniques. Morgan Kaufmann, 2000.T. Imielinski and H. Mannila. A database perspective on knowledge discovery. Communications of ACM, 39:58-64, 1996.G. Piatetsky-Shapiro, U. Fayyad, and P. Smith. From data mining to knowledge discovery: An overview. In U.M. Fayyad, et al. (eds.), Advances in Knowledge Discovery and Data Mining, AAAI/MIT Press, 1996.G. Piatetsky-Shapiro and W. J. Frawley. Knowledge Discovery in Databases. AAAI/MIT Press, 1991.
Your consent to our cookies if you continue to use this website.