CMPUT 391 Database Management Systems. Data Mining. Textbook: Chapter (without 17.10)

Transcription

1 CMPUT 391 Database Management Systems Data Mining Textbook: Chapter (without 17.10) University of Alberta 1

2 Overview Motivation KDD and Data Mining Association Rules Clustering Classification University of Alberta 2

3 The Current Situation Data Rich Technology is available to help us collect data Bar code, scanners, satellites, cameras, etc. Technology is available to help us store data Databases, data warehouses, variety of repositories Information Poor We need powerful data analysis processes to turn the data into information (competitive edge, research, etc.) Sales / Service Transactions Sky and Earth Observation Molecular Biology Medical Data... University of Alberta 3

4 Overview Motivation KDD and Data Mining Association Rules Clustering Classification University of Alberta 4

5 KDD: Knowledge Discovery in Databases Confluence of Multiple Disciplines Machine Learning Statistics Database Technology KDD Visualization... Process of non trivial extraction of implicit, previously unknown and potentially useful information from large collections of data University of Alberta 5

6 The KDD Process Data mining: the core of knowledge discovery process. Pattern Evaluation Task-relevant Data Data Cleaning Data Warehouse Selection and Transformation KDD is an Iterative Process Data Integration University of Alberta 6

7 Basic Data Mining Functionalities Association Rules Find frequent associations in transaction databases Clustering Find natural groups in the data Classification Find models to predict class membership for new objects Concept Characterization and Discrimination Generalize, summarize, and contrast data characteristics Other methods Outlier detection Sequential patterns Methods for special data types, e.g., spatial data mining, web mining, text mining, University of Alberta 7

8 Overview Motivation KDD and Data Mining Association Rules Clustering Classification University of Alberta 8

9 Basic Concepts I ={i 1, i 2, i n } is the set of all possible items A transaction is a set of items: T={i a, i b, i t }, T I D, is a set (database) of transactions the task relevant data. An association rule is of the form: P Q where P I, Q I, and P Q = University of Alberta 9

10 Rule Measures: Support and Confidence Support of a rule P Q - s D (P Q) = s D (P Q): Probability that a transaction in D contains the items from both itemsets P and Q. Customer buys both Customer buys bread Confidence of a rule P Q - c D (P Q) = s D (P Q) / s D (P): Probability that a transaction in D that already contains the items in P also contains the items in Q. Determined by counting the corresponding relative frequencies in the database Customer buys milk University of Alberta 10

11 Support and Confidence An Example Transaction ID Items Bought 2000 A,B,C 1000 A,C 4000 A,D 5000 B,E,F For rule A C: support = support({a, C}) = 50% Min. support 50% Min. confidence 50% Frequent Itemset Support {A} 75% {B} 50% {C} 50% {A,C} 50% confidence = support({a, C})/support({A}) = 66.6% University of Alberta 11

12 Strong Rules Thresholds: minimum support: minsup minimum confidence: minconf Itemsets are called frequent if their support is greater or equal than minsup Task: Find all rules P Q [s, c] so that s D (P Q) minsup c D (P Q) minconf University of Alberta 12

13 Mining Association Rules Input A database of transactions Each transaction is a list of items (Ex. purchased by a customer in a visit) Task: Find Find all rules P Q [s, c] that associate the presence of one set of items with that of another set of items so that sd(p Q) minsup cd(p Q) minconf Example: 98% of people who purchase tires and auto accessories also get automotive services done University of Alberta 13

14 Mining Frequent Itemsets: the Key Step Iteratively find the frequent itemsets, i.e. sets of items that have minimum support, with cardinality from 1 to k (k-itemsets) Based on the Apriori principle: Any subset of a frequent itemset must also be frequent. E.g., if {AB} is a frequent itemset, both {A} and {B} must be frequent itemsets. Method based on the apriori principle When generating candidate k-itemsets for the next iteration, only consider those where all subsets of length k -1 have been determined as frequent in the previous step Use the frequent itemsets to generate association rules. University of Alberta 14

15 The Apriori Algorithm Join Step: C k is generated by joining L k-1 with itself Prune Step: Any (k-1)-itemset that is not frequent cannot be a subset of a frequent k-itemset Pseudo-code: C k : Candidate itemset of size k L k : frequent itemset of size k L 1 = {frequent items}; for (k = 1; L k!= ; k++) do begin C k+1 = candidates generated from L k ; for each transaction t in database do increment the count of all candidates in C k+1 that are contained in t L k+1 = candidates in C k+1 with min_support end return k L k ; University of Alberta 15

16 Candidate Generation: 1. Join Step Suppose the items in L k-1 are listed in an order Step 1: self-joining L k-1 insert into C k select p.item 1, p.item 2,, p.item k-1, q.item k-1 from L k-1 p, L k-1 q where p.item 1 =q.item 1,, p.item k-2 =q.item k-2, p.item k-1 < q.item k-1 p L k-1 (1, 2, 3) q L k-1 (1, 2, 4) (1, 2, 3, 4) C k University of Alberta 16

17 Candidate Generation: 2. Prune Step Step 2: pruning forall itemsets c in C k do forall (k-1)-subsets s of c do if (s is not in L k-1 ) then delete c from C k Example L 3 = {(1 2 3), (1 2 4), (1 3 4), (1 3 5), (2 3 4)} Candidates after the join step: {( ), ( )} In the pruning step: delete ( ) because (3 4 5) L 3 C 4 = {( )} University of Alberta 17

18 The Apriori Algorithm -- Example Database D TID Items L 2 itemset sup {1 3} 2 {2 3} 2 {2 5} 3 {3 5} 2 C 3 itemset {2 3 5} C 1 Scan D C 2 itemset sup. {1} 2 {2} 3 {3} 3 {4} 1 {5} 3 itemset sup {1 2} 1 {1 3} 2 {1 5} 1 {2 3} 2 {2 5} 3 {3 5} 2 Scan D L 3 L 1 Scan D itemset sup {2 3 5} 2 itemset sup. {1} 2 {2} 3 {3} 3 {5} 3 C 2 itemset {1 2} {1 3} {1 5} {2 3} {2 5} {3 5} minsup = 50% Note: {1,2,3}{1,2,5} and {1,3,5} not in C 3 University of Alberta 18

19 Generating Association Rules from Frequent Itemsets Only strong association rules are generated. Frequent itemsets satisfy minimum support threshold. Strong AR satisfy minimum confidence threshold. Confidence(P Q) = Prob(Q/P) = Support(P Q) Support(P) For each frequent itemset, f, generate all non-empty subsets of f. For every non-empty subset s of f do output rule s (f-s) if support(f)/support(s) min_confidence end University of Alberta 19

20 Interestingness of Association Rules Motivation Example (Aggarwal & Yu, PODS98) Among 5000 students 3000 play basketball 3750 eat cereal 2000 both play basket ball and eat cereal play basketball eat cereal [40%, 66.7%] is misleading because the overall percentage of students eating cereal is 75% which is higher than 66.7%. play basketball not eat cereal [20%, 33.3%] is far more accurate, although with lower support and confidence basketball not basketball sum(row) cereal not cereal sum(col.) University of Alberta 20

21 Interestingness of Association Rules Delete misleading association rules Condition for a rule A B P( A B) P( A) > P( B) + Measure for the interestingness of a rule P( A B) P( B) P( A) The larger the value, the more interesting the relation between A and B, expressed by the rule. Other measures: correlation between A and B d for a suitable threshold d > 0 University of Alberta 21

22 Overview Motivation KDD and Data Mining Association Rules Clustering Classification University of Alberta 22

23 What is Clustering in Data Mining? Clustering is a process of partitioning a set of data (or objects) in a set of meaningful sub-classes, called clusters. Helps users understand the natural grouping or structure in a data set. Cluster: a collection of data objects that are similar to one another and thus can be treated collectively as one group. Clustering: unsupervised classification: no predefined classes. University of Alberta 23

24 What Is Good Clustering? A good clustering method will produce high quality clusters where: the intra-class similarity (that is within a cluster) is high. the inter-class similarity (that is between clusters) is low. The quality of a clustering result also depends on the similarity measure used by the method. The quality of a clustering result also depends on the definition and representation of cluster different clustering algorithms may have different underlying notions of clusters. University of Alberta 24

25 Major Clustering Techniques Partitioning algorithms: Construct various partitions and then evaluate them by some criterion. Hierarchical algorithms: Create a hierarchical decomposition of the set of data (or objects) using some criterion. There is an agglomerative approach and a divisive approach. Density-based: based on connectivity and density functions. Grid-based: based on a multiple-level granularity structure. Model-based: A model is hypothesized for each of the clusters and the idea is to find the best fit of that model to each other. University of Alberta 25

26 Partitioning Algorithms: Basic Concept Partitioning method: Given a number k, partition a database D of n objects into a set of k clusters so that a chosen objective function is minimized (e.g., sum of distances to the center of the clusters). Global optimum: exhaustively enumerate all partitions too expensive! Heuristic methods based on iterative refinement of an initial partition Bad Clustering Optimal Clustering x x x x Centroids x x x x Centroids University of Alberta 26

27 The K-Means Clustering Method Given k, the k-means algorithm is implemented in 4 steps: 1. Partition objects into k nonempty subsets 2. Compute seed points as the centroids of the clusters of the current partition. The centroid of a cluster for the k-means algorithm is the mean point of all points in the cluster. 3. Assign each object to the cluster with the nearest seed point. 4. Go back to Step 2, stop when no more new assignment University of Alberta 27

28 Comments on the K-Means Method Strength of the k-means: Relatively efficient: O(tkn), where n is # of objects, k is # of clusters, and t is # of iterations. Normally, k, t << n. Weakness of the k-means: Applicable only when mean is defined, then what about categorical data? Need to specify k, the number of clusters, in advance. Not suitable to discover clusters with non-convex shapes. Often terminates at a local optimum. University of Alberta 28

29 Hierarchical Clustering Hierarchical decomposition of the data set (with respect to a given similarity measure) into a set of nested clusters Result represented by a so called dendrogram Nodes in the dendrogram represent possible clusters can be constructed bottom-up (agglomerative approach) or top down (divisive approach) Step 0 Step 1 Step 2 Step 3 Step 4 a b c d e a b d e a b c d e c d e Step 4 Step 3 Step 2 Step 1 Step 0 agglomerative divisive University of Alberta 29

30 Hierarchical Clustering: Example Interpretation of the dendrogram The root represents the whole data set A leaf represents a single objects in the data set An internal node represent the union of all objects in its sub-tree The height of an internal node represents the distance/similarity between its two child nodes distance between groups University of Alberta 30

31 Agglomerative Hierarchical Clustering Single-Link Method and Variants: start by placing each object in its own cluster. keep merging closest pairs (most similar pairs) of clusters into larger clusters until all objects are in a single cluster. Most hierarchical methods belong to this category. They differ mainly in their definition of betweencluster similarity. Single-Link: similarity is defined as the similarity between the closest (i.e., most similar) pair of objects. University of Alberta 31

32 Overview Motivation KDD and Data Mining Association Rules Clustering Classification University of Alberta 32

33 What is Classification? The goal of data classification is to organize and categorize data in distinct classes. A model is first created based on the data distribution. The model is then used to classify new data. Given the model, a class can be predicted for new data.? n University of Alberta 33

34 Classification is a three-step process 1. Model construction (Learning): Each tuple is assumed to belong to a predefined class, as determined by one of the attributes, called the class label. Training Set: Outlook Tempreature Humidity W indy Class sunny hot high false N sunny hot high true N overcast hot high false P The set of all tuples used for construction of the model is called training set. The model can be represented in the following forms: Classification rules, (IF-THEN statements), Decision tree Mathematical formulae University of Alberta 34

35 1. Classification Process (Learning) Training Data Classification Algorithms Name Income Age Credit rating Bruce Low <30 bad Dave Medium [30..40] good William High <30 good Marie Medium >40 good Anne Low [30..40] good Chris Medium <30 bad Classifier (Model) IF Income = High OR Age > 30 THEN CreditRating = Good University of Alberta 35

36 Classification is a three-step process 2. Model Evaluation (Accuracy): Test Set: Estimate accuracy rate of the model based on a test set. The known label of test sample is compared with the classified result from the model. Accuracy rate is the percentage of test set samples that are correctly classified by the model. Test set is independent of training set otherwise over-fitting will occur. Outlook Tempreature Humidity W indy Class rainy hot low false P sunny hot high true N overcast hot high false P = Predicted Class? = P? = P? = N? University of Alberta 36

37 2. Classification Process (Accuracy Evaluation) Testing Data Classifier (Model) Name Income Age Credit rating Tom Medium <30 bad Jane High <30 bad Wei High >40 good Hua Medium [30..40] good How accurate is the model? IF Income = High OR Age > 30 THEN CreditRating = Good University of Alberta 37

38 Classification is a three-step process 3. Model Use (Classification): The model is used to classify new objects where the class label is not known. Using the attributes of the new object and the model, assign a class label to the new object Outlook Tempreature Humidity W indy Class New Data: rainy hot low false? Predict Class! rainy mild high false? overcast hot high true? University of Alberta 38

39 3. Classification Process (Classification) New Data Classifier (Model) Name Income Age Credit rating Paul High [30..40]? Credit Rating? University of Alberta 39

40 Classification Methods Decision Tree Induction Neural Networks Bayesian Classification Association-Based Classification K-Nearest Neighbour Genetic Algorithms Etc. University of Alberta 40

41 What is a Decision Tree? Atr=? CL A decision tree is a flow-chart-like tree structure. Internal node denotes a test on an attribute Branch represents an outcome of the test All tuples in branch have the same value for the tested attribute. Leaf node represents class label or class label Atr=? distribution. Atr=? Atr=? Atr=? CL CL CL CL CL CL CL University of Alberta 41

42 Training Dataset An Example from Quinlan s ID3 Outlook Tempreature Humidity W indy Class sunny hot high false N sunny hot high true N overcast hot high false P rain mild high false P rain cool normal false P rain cool normal true N overcast cool normal true P sunny mild high false N sunny cool normal false P rain mild normal false P sunny mild normal true P overcast mild high true P overcast hot normal false P rain mild high true N University of Alberta 42

43 A Sample Decision Tree sunny Humidity? Outlook? overcast P rain Windy? Outlook Tempreature Humidity Windy Class sunny hot high false N sunny hot high true N overcast hot high false P rain mild high false P rain cool normal false P rain cool normal true N overcast cool normal true P sunny mild high false N sunny cool normal false P rain mild normal false P sunny mild normal true P overcast mild high true P overcast hot normal false P rain mild high true N high normal true false N P N P University of Alberta 43

44 Decision-Tree Classification Methods The basic top-down decision tree generation approach usually consists of two phases: 1. Tree construction At the start, all the training examples are at the root. Partition examples recursively, based on selected attributes. 2. Tree pruning Aiming at removing tree branches that may reflect noise in the training data and lead to errors when classifying test data improve classification accuracy. University of Alberta 44

45 Decision Tree Construction Atr=? Recursive process: Tree starts a single node representing all data. Recursion stops when: a) Sample in node belong to the same class; b) There are no remaining attributes on which to split; a) & b) node becomes a leaf labeled with the majority class label. There are no samples with attribute value. CL Otherwise, select suitable attribute partition the data according to the attribute values of the selected attribute into subsets. For each of these subsets: create a new child node under the current parent node and recursively apply the method to the new child nodes. University of Alberta 45

46 Partitioning the Data at a Given Node Split criterion: Use a goodness/impurity function to determine the attribute that results in the purest subsets with respect to the class label. Different goodness functions exist: information gain, gini index, etc. Branching scheme: binary splitting (numerical attributes, gini index) versus many splitting (categorical attributes, information gain). University of Alberta 46

47 Example for Algorithm (ID3) All attributes are categorical Create a node N; if samples are all of the same class C, then return N as a leaf node labeled with C. if attribute-list is empty then return N as a leaf node labeled with the most common class. Select split-attribute with highest information gain label N with the split-attribute for each value A i of split-attribute, grow a branch from Node N let S i be the branch in which all tuples have the value A i for split- attribute if S i is empty then attach a leaf labeled with the most common class. Else recursively run the algorithm at Node S i Until all branches reach leaf nodes University of Alberta 47

48 How to use a tree? Directly test the attribute values of an unknown sample against the tree. A path is traced from root to a leaf which holds the label. Indirectly decision tree is converted to classification rules. one rule is created for each path from the root to a leaf. IF-THEN rules are easier for humans to understand. University of Alberta 48