chapter 6. classification and prediction 數(shù)據(jù)挖掘：概念與技術(shù)--ppt 英文版

1、April 2, 2024,Data Mining: Concepts and Techniques,1,Data Mining: Concepts and Techniques — Chapter 6 —,Jiawei HanDepartment of Computer Science University of Illinois at Urbana-Champaignwww.cs.uiuc.edu/~hanj

2、9;2006 Jiawei Han and Micheline Kamber, All rights reserved,April 2, 2024,Data Mining: Concepts and Techniques,2,Chapter 6. Classification and Prediction,What is classification? What is prediction?Issues regarding class

3、ification and predictionClassification by decision tree inductionBayesian classificationRule-based classificationClassification by back propagation,Support Vector Machines (SVM) Associative classification Lazy lear

4、ners (or learning from your neighbors)Other classification methodsPredictionAccuracy and error measuresEnsemble methodsModel selectionSummary,,April 2, 2024,Data Mining: Concepts and Techniques,3,Classification p

5、redicts categorical class labels (discrete or nominal)classifies data (constructs a model) based on the training set and the values (class labels) in a classifying attribute and uses it in classifying new dataPredictio

6、n models continuous-valued functions, i.e., predicts unknown or missing values Typical applicationsCredit approvalTarget marketingMedical diagnosisFraud detection,Classification vs. Prediction,April 2, 2024,Data M

7、ining: Concepts and Techniques,4,Classification—A Two-Step Process,Model construction: describing a set of predetermined classesEach tuple/sample is assumed to belong to a predefined class, as determined by the class la

8、bel attributeThe set of tuples used for model construction is training setThe model is represented as classification rules, decision trees, or mathematical formulaeModel usage: for classifying future or unknown object

9、sEstimate accuracy of the modelThe known label of test sample is compared with the classified result from the modelAccuracy rate is the percentage of test set samples that are correctly classified by the modelTest se

10、t is independent of training set, otherwise over-fitting will occurIf the accuracy is acceptable, use the model to classify data tuples whose class labels are not known,April 2, 2024,Data Mining: Concepts and Techniques

11、,5,Process (1): Model Construction,TrainingData,,,ClassificationAlgorithms,,IF rank = ‘professor’OR years > 6THEN tenured = ‘yes’,Classifier(Model),,,,April 2, 2024,Data Mining: Concepts and Techniques,6,Process

12、(2): Using the Model in Prediction,Classifier,TestingData,,,,,Unseen Data,(Jeff, Professor, 4),,,,Tenured?,April 2, 2024,Data Mining: Concepts and Techniques,7,Supervised vs. Unsupervised Learning,Supervised learning (c

13、lassification)Supervision: The training data (observations, measurements, etc.) are accompanied by labels indicating the class of the observationsNew data is classified based on the training setUnsupervised learning (

14、clustering)The class labels of training data is unknownGiven a set of measurements, observations, etc. with the aim of establishing the existence of classes or clusters in the data,April 2, 2024,Data Mining: Concepts a

15、nd Techniques,8,Chapter 6. Classification and Prediction,What is classification? What is prediction?Issues regarding classification and predictionClassification by decision tree inductionBayesian classificationRule-b

16、ased classificationClassification by back propagation,Support Vector Machines (SVM) Associative classification Lazy learners (or learning from your neighbors)Other classification methodsPredictionAccuracy and error

17、 measuresEnsemble methodsModel selectionSummary,,April 2, 2024,Data Mining: Concepts and Techniques,9,Issues: Data Preparation,Data cleaningPreprocess data in order to reduce noise and handle missing valuesRelevance

18、 analysis (feature selection)Remove the irrelevant or redundant attributesData transformationGeneralize and/or normalize data,April 2, 2024,Data Mining: Concepts and Techniques,10,Issues: Evaluating Classification Met

19、hods,Accuracyclassifier accuracy: predicting class labelpredictor accuracy: guessing value of predicted attributesSpeedtime to construct the model (training time)time to use the model (classification/prediction time

20、)Robustness: handling noise and missing valuesScalability: efficiency in disk-resident databases Interpretabilityunderstanding and insight provided by the modelOther measures, e.g., goodness of rules, such as decisi

21、on tree size or compactness of classification rules,April 2, 2024,Data Mining: Concepts and Techniques,11,Chapter 6. Classification and Prediction,What is classification? What is prediction?Issues regarding classificati

22、on and predictionClassification by decision tree inductionBayesian classificationRule-based classificationClassification by back propagation,Support Vector Machines (SVM) Associative classification Lazy learners (o

23、r learning from your neighbors)Other classification methodsPredictionAccuracy and error measuresEnsemble methodsModel selectionSummary,,April 2, 2024,Data Mining: Concepts and Techniques,12,Decision Tree Induction:

24、 Training Dataset,This follows an example of Quinlan’s ID3 (Playing Tennis),April 2, 2024,Data Mining: Concepts and Techniques,13,Output: A Decision Tree for “buys_computer”,April 2, 2024,Data Mining: Concepts and Techn

25、iques,14,Algorithm for Decision Tree Induction,Basic algorithm (a greedy algorithm)Tree is constructed in a top-down recursive divide-and-conquer mannerAt start, all the training examples are at the rootAttributes are

26、 categorical (if continuous-valued, they are discreted in advance)Samples are partitioned recursively based on selected attributesTest attributes are selected on the basis of a heuristic or statistical measure (e.g., i

27、nformation gain)Conditions for stopping partitioningAll samples for a given node belong to the same classThere are no remaining attributes for further partitioning – majority voting is employed for classifying the lea

28、fThere are no samples left,April 2, 2024,Data Mining: Concepts and Techniques,15,Attribute Selection Measure: Information Gain (ID3/C4.5),Select the attribute with the highest information gainLet pi be the probability

29、that an arbitrary tuple in D belongs to class Ci, estimated by |Ci, D|/|D|Expected information (entropy) needed to classify a tuple in D:Information needed (after using A to split D into v partitions) to classify D:

30、Information gained by branching on attribute A,April 2, 2024,Data Mining: Concepts and Techniques,16,Attribute Selection: Information Gain,Class P: buys_computer = “yes”Class N: buys_computer = “no”,means “age <=30”

31、has 5 out of 14 samples, with 2 yes’es and 3 no’s. HenceSimilarly,,April 2, 2024,Data Mining: Concepts and Techniques,17,Computing Information-Gain for Continuous-Value Attributes,Let attribute A be a continuous-va

32、lued attributeMust determine the best split point for ASort the value A in increasing orderTypically, the midpoint between each pair of adjacent values is considered as a possible split point(ai+ai+1)/2 is the midpoi

33、nt between the values of ai and ai+1The point with the minimum expected information requirement for A is selected as the split-point for ASplit:D1 is the set of tuples in D satisfying A ≤ split-point, and D2 is the se

34、t of tuples in D satisfying A > split-point,April 2, 2024,Data Mining: Concepts and Techniques,18,Gain Ratio for Attribute Selection (C4.5),Information gain measure is biased towards attributes with a large number of

35、valuesC4.5 (a successor of ID3) uses gain ratio to overcome the problem (normalization to information gain)GainRatio(A) = Gain(A)/SplitInfo(A)Ex.gain_ratio(income) = 0.029/0.926 = 0.031The attribute with the maxim

36、um gain ratio is selected as the splitting attribute,April 2, 2024,Data Mining: Concepts and Techniques,19,Gini index (CART, IBM IntelligentMiner),If a data set D contains examples from n classes, gini index, gini(D) is

37、defined as where pj is the relative frequency of class j in DIf a data set D is split on A into two subsets D1 and D2, the gini index gini(D) is defined asReduction in Impurity:The attribute provides t

38、he smallest ginisplit(D) (or the largest reduction in impurity) is chosen to split the node (need to enumerate all the possible splitting points for each attribute),April 2, 2024,Data Mining: Concepts and Techniques,20,G

39、ini index (CART, IBM IntelligentMiner),Ex. D has 9 tuples in buys_computer = “yes” and 5 in “no”Suppose the attribute income partitions D into 10 in D1: {low, medium} and 4 in D2but gini{medium,high} is 0.30 and

40、thus the best since it is the lowestAll attributes are assumed continuous-valuedMay need other tools, e.g., clustering, to get the possible split valuesCan be modified for categorical attributes,April 2, 2024,Data Min

41、ing: Concepts and Techniques,21,Comparing Attribute Selection Measures,The three measures, in general, return good results butInformation gain: biased towards multivalued attributesGain ratio: tends to prefer unbalan

42、ced splits in which one partition is much smaller than the othersGini index: biased to multivalued attributeshas difficulty when # of classes is largetends to favor tests that result in equal-sized partitions and pur

43、ity in both partitions,April 2, 2024,Data Mining: Concepts and Techniques,22,Other Attribute Selection Measures,CHAID: a popular decision tree algorithm, measure based on χ2 test for independenceC-SEP: performs better t

44、han info. gain and gini index in certain casesG-statistics: has a close approximation to χ2 distribution MDL (Minimal Description Length) principle (i.e., the simplest solution is preferred): The best tree as the one

45、that requires the fewest # of bits to both (1) encode the tree, and (2) encode the exceptions to the treeMultivariate splits (partition based on multiple variable combinations)CART: finds multivariate splits based on a

46、 linear comb. of attrs.Which attribute selection measure is the best? Most give good results, none is significantly superior than others,April 2, 2024,Data Mining: Concepts and Techniques,23,Overfitting and Tree Prunin

47、g,Overfitting: An induced tree may overfit the training data Too many branches, some may reflect anomalies due to noise or outliersPoor accuracy for unseen samplesTwo approaches to avoid overfitting Prepruning: Halt

48、 tree construction early—do not split a node if this would result in the goodness measure falling below a thresholdDifficult to choose an appropriate thresholdPostpruning: Remove branches from a “fully grown” tree—get

49、a sequence of progressively pruned treesUse a set of data different from the training data to decide which is the “best pruned tree”,April 2, 2024,Data Mining: Concepts and Techniques,24,Enhancements to Basic Decision T

50、ree Induction,Allow for continuous-valued attributesDynamically define new discrete-valued attributes that partition the continuous attribute value into a discrete set of intervalsHandle missing attribute valuesAssign

51、 the most common value of the attributeAssign probability to each of the possible valuesAttribute constructionCreate new attributes based on existing ones that are sparsely representedThis reduces fragmentation, repe

52、tition, and replication,,,,April 2, 2024,Data Mining: Concepts and Techniques,25,Classification in Large Databases,Classification—a classical problem extensively studied by statisticians and machine learning researchers

53、Scalability: Classifying data sets with millions of examples and hundreds of attributes with reasonable speedWhy decision tree induction in data mining?relatively faster learning speed (than other classification method

54、s)convertible to simple and easy to understand classification rulescan use SQL queries for accessing databasescomparable classification accuracy with other methods,April 2, 2024,Data Mining: Concepts and Techniques,26

55、,Scalable Decision Tree Induction Methods,SLIQ (EDBT’96 — Mehta et al.)Builds an index for each attribute and only class list and the current attribute list reside in memorySPRINT (VLDB’96 — J. Shafer et al.)Construct

56、s an attribute list data structure PUBLIC (VLDB’98 — Rastogi & Shim)Integrates tree splitting and tree pruning: stop growing the tree earlierRainForest (VLDB’98 — Gehrke, Ramakrishnan & Ganti)Builds an AVC-li

57、st (attribute, value, class label)BOAT (PODS’99 — Gehrke, Ganti, Ramakrishnan & Loh)Uses bootstrapping to create several small samples,April 2, 2024,Data Mining: Concepts and Techniques,27,Scalability Framework for

58、 RainForest,Separates the scalability aspects from the criteria that determine the quality of the tree Builds an AVC-list: AVC (Attribute, Value, Class_label) AVC-set (of an attribute X )Projection of training datase

59、t onto the attribute X and class label where counts of individual class label are aggregatedAVC-group (of a node n )Set of AVC-sets of all predictor attributes at the node n,April 2, 2024,Data Mining: Concepts and Tec

60、hniques,28,Rainforest: Training Set and Its AVC Sets,AVC-set on income,AVC-set on Age,AVC-set on Student,Training Examples,AVC-set on credit_rating,April 2, 2024,Data Mining: Concepts and Techniques,29,Data Cube-Based

61、Decision-Tree Induction,Integration of generalization with decision-tree induction (Kamber et al.’97)Classification at primitive concept levelsE.g., precise temperature, humidity, outlook, etc.Low-level concepts, scat

62、tered classes, bushy classification-treesSemantic interpretation problemsCube-based multi-level classificationRelevance analysis at multi-levelsInformation-gain analysis with dimension + level,April 2, 2024,Data Mini

63、ng: Concepts and Techniques,30,BOAT (Bootstrapped Optimistic Algorithm for Tree Construction),Use a statistical technique called bootstrapping to create several smaller samples (subsets), each fits in memoryEach subset

64、is used to create a tree, resulting in several trees These trees are examined and used to construct a new tree T’It turns out that T’ is very close to the tree that would be generated using the whole data set together

65、Adv: requires only two scans of DB, an incremental alg.,April 2, 2024,Data Mining: Concepts and Techniques,31,Presentation of Classification Results,April 2, 2024,Data Mining: Concepts and Techniques,32,Visualization of

66、a Decision Tree in SGI/MineSet 3.0,April 2, 2024,Data Mining: Concepts and Techniques,33,Interactive Visual Mining by Perception-Based Classification (PBC),April 2, 2024,Data Mining: Concepts and Techniques,34,Chapter 6.

67、 Classification and Prediction,What is classification? What is prediction?Issues regarding classification and predictionClassification by decision tree inductionBayesian classificationRule-based classificationClassi

68、fication by back propagation,Support Vector Machines (SVM) Associative classification Lazy learners (or learning from your neighbors)Other classification methodsPredictionAccuracy and error measuresEnsemble methods