The k-means Algorithm and Genetic Algorithm

Transcription

1 The k-means Algorithm and Genetic Algorithm

2 k-means algorithm Genetic algorithm Rough set approach Fuzzy set approaches Chapter 8 2

3 The K-Means Algorithm The K-Means algorithm is a simple yet effective statistical clustering technique. Here is the algorithm: 1. Choose a value for K, the total number of clusters to be determined. 2. Choose K instances (data points) within the dataset at random. These are the initial cluster centers. 3. Use simple Euclidean distance to assign the remaining instances to their closest cluster center. Chapter 8 3

4 The K-Means Algorithm 4. Use the instances in each cluster to calculate a new mean for each cluster. 5. If the new mean values are identical to the mean values of the previous iteration the process terminates. Otherwise, use the new means as cluster centers and repeat steps 3-5. Chapter 8 4

5 The K-Means Algorithm An Example Using K-Means Chapter 8 5

6 The K-Means Algorithm An Example Using K-Means Chapter 8 6

7 The K-Means Algorithm General Considerations Chapter 8 7

8 The K-Means Algorithm General Considerations Chapter 8 8

9 All instances correspond to points in the n-d space. The nearest neighbor are defined in terms of Euclidean distance. The target function could be discrete- or realvalued _ xq _ Chapter 8 9

10 For discrete-valued, the k-nn returns the most common value among the k training examples nearest to xq. Vonoroi diagram: the decision surface induced by 1-NN for a typical set of training examples _ xq _ Chapter 8 10

11 The k-nn algorithm for continuous-valued target functions Calculate the mean values of the k nearest neighbors Distance-weighted nearest neighbor algorithm Weight the contribution of each of the k neighbors according to their distance to the query point x q giving greater weight to closer neighbors Similarly, for real-valued 1 target functions w d ( xq, x i ) 2 Chapter 8 11

12 Genetic Learning Here we present a basic genetic learning algorithm. 1. Initialize a population P of n elements, often referred to as chromosomes, as a potential solution. 2. Until a specified termination condition is satisfied: a. Use a fitness function to evaluate each element of the current solution. If an element passes the fitness criteria, it remains in P. b. The population now contains m elements (m<=n). Use genetic operators to create (n-m) new elements. Add the new elements to the population. Chapter 8 12

13 Genetic Learning Genetic Algorithms and Supervised Learning Chapter 8 13

14 Genetic Learning Genetic Algorithms and Supervised Learning Chapter 8 14

15 Genetic Learning Genetic Algorithms and Supervised Learning Chapter 8 15

16 Genetic Learning Genetic Algorithms and Supervised Learning Chapter 8 16

17 Genetic Learning Genetic Algorithms and... Supervised Learning Chapter 8 17

18 Genetic Learning Genetic Algorithms and..unsupervised Clustering Chapter 8 18

19 Genetic Learning Genetic Algorithms and Unsupervised Clustering Chapter 8 19

20 Genetic Learning General Considerations Here is a list of considerations when using a problem-solving approach based on genetic learning: Genetic algorithms are designed to find globally optimized solutions. However, there is no guarantee that any given solution is not the result of a local rather than a global optimization. The fitness function determines the computational complexity of a genetic algorithm. A fitness function involving several calculations can be computationally expensive. Chapter 8 20

21 Genetic Learning General Considerations Genetic algorithms explain their results to the extent that the fitness function is understandable. Transforming the data to form suitable for a genetic algorithm can be a challenge. Chapter 8 21

22 GA: based on an analogy to biological evolution Each rule is represented by a string of bits An initial population is created consisting of randomly generated rules Based on the notion of survival of the fittest, a new population is formed to consists of the fittest rules and their offsprings The fitness of a rule is represented by its classification accuracy on a set of training examples Offsprings are generated by crossover and mutation Chapter 8 22

23 Population-based technique for discovery of...knowledge structures Based on idea that evolution represents search for optimum solution set Massively parallel Chapter 8 23

24 Population Set of individuals, each represented by one or more strings of characters Chromosome The string representing an individual Chapter 8 24

25 Chromosome Gene (Allele="0") Locus=5 Gene The basic informational unit on a chromosome Allele :The value of a specific gene Locus : The ordinal place... on a chromosome where a specific gene is found Chapter 8 25

26 Reproduction Increase representations of strong individuals Crossover Explore the search space Mutation Recapture lost genes due to crossover Chapter 8 26

27 Parent 1: Parent 2: Simple reproduction Offspring 1: Offspring 2: Parent 1: Parent 2: Reproduction with Offspring crossover at locus 3 1: Offspring 2: Parent 1: Parent 2: Offspring 1: Simple reproduction with mutation at locus 3 for offspring 1 Offspring 2: Chapter 8 27

28 Ability of an individual to survive into the next generation Survival of the fittest Usually calculated in terms of an objective fitness function Maximization Minimization Other functions Chapter 8 28

29 Based on adaptation and evolution Structures undergoing adaptation are computer programs of varying size and shape Computer programs are genetically bred over time Chapter 8 29

30 Rule-based knowledge discovery and concept learning tool Operates by means of evaluation, credit assignment, and discovery applied to a population of chromosomes (rules) each with a corresponding phenotype (outcome) Chapter 8 30

31 Performance Provides interaction between environment and rule base Performs matching function Reinforcement Rewards accurate classifiers Punishes inaccurate classifiers Discovery Uses the genetic algorithm to search for plausible rules Chapter 8 31

32 Rough sets are used to approximately or roughly define equivalent classes A rough set for a given class C is approximated by two sets: a lower approximation (certain to be in C) an upper approximation (cannot be described as not belonging to C) Chapter 8 32

33 Fuzzy logic uses truth values between 0.0 and 1.0 to represent the degree of membership (such as using fuzzy membership graph) Attribute values are converted to fuzzy values e.g., income is mapped into the discrete categories {low, medium, high} with fuzzy values calculated Chapter 8 33

34 For a given new sample, more than one fuzzy value may apply Each applicable rule contributes a vote for membership in the categories Typically, the truth values for each predicted category are summed. Chapter 8 34

35 Chapter Summary The K-Means algorithm is a statistical unsupervised clustering technique. All input attributes to the algorithm must be numeric and the user is required to make a decision about... how many clusters are to be discovered. The algorithm begins by randomly choosing one data point to represent each cluster. Each data instance is then placed in the cluster to which it is most similar. New cluster centers are computed and the process continues until...the cluster centers do not change. Chapter 8 35

36 Chapter Summary The K-Means algorithm is easy to implement and understand. However, the algorithm is not guaranteed to converge to a globally optimal solution, lacks the ability to explain what has been found, unable to tell which attributes are significant in determining the formed clusters. Despite these limitations, the K-Means algorithm is among the most widely used clustering techniques. Chapter 8 36

37 Chapter Summary Genetic algorithms apply the theory of evolution to inductive learning. Genetic learning can be supervised...or...unsupervised typically used for problems that cannot be solved with traditional techniques. A standard genetic approach to learning applies a fitness function to a set of data elements to determine... which elements survive from one generation to the next. Chapter 8 37

38 Chapter Summary Those elements not surviving are used to create new instances to replace deleted elements. In addition to being used for supervised learning and unsupervised clustering, genetic techniques can be employed in conjunction with other learning techniques. Chapter 8 38

39 Key Terms Affinity analysis. The process of determining which things are typically grouped together. Confidence. Given a rule of the form If A then B, confidence is defined as the conditional probability that B is true when A is known to be true. Crossover. A genetic learning operation that creates new population elements by combining parts of two or more elements from the current population. Chapter 8 39

40 Key Terms Genetic algorithm. A data mining technique based on the theory of evolution. Mutation. A genetic learning operation that creates a new population element by randomly modifying a portion of an existing element. Selection. A genetic learning operation that adds copies of current population elements with high fitness scores to the next generation of the population. Chapter 8 40

41 Data Mining: Concepts and Techniques (Chapter 7 Slide for textbook), Jiawei Han and Micheline Kamber, Intelligent Database Systems Research Lab, School of Computing Science, Simon Fraser University, Canada Chapter 8 41