This preview shows page 1. Sign up to view the full content.
Unformatted text preview: Clustering Algorithms
Hierarchical Clustering k Means Algorithms CURE Algorithm
1 Methods of Clustering
x Hierarchical (Agglomerative): Initially, each point in cluster by itself. Repeatedly combine the two "nearest" clusters into one. Maintain a set of clusters. Place points into their "nearest" cluster.
2 x Point Assignment: Hierarchical Clustering
x Two important questions:
1. How do you determine the "nearness" of clusters? 2. How do you represent a cluster of more than one point? 3 Hierarchical Clustering (2)
x Key problem: as you build clusters, how do you represent the location of each cluster, to tell which pair of clusters is closest? x Euclidean case: each cluster has a centroid = average of its points. Measure intercluster distances by distances of centroids. 4 Example (5,3) o (1,2) x (1.5,1.5) x (1,1) o (2,1) o (0,0) o x (4.7,1.3) o (4,1) x (4.5,0.5) o (5,0)
5 And in the NonEuclidean Case?
x The only "locations" we can talk about are the points themselves. x Approach 1: clustroid = point "closest" to other points. I.e., there is no "average" of two points. Treat clustroid as if it were centroid, when computing intercluster distances. 6 "Closest" Point?
x Possible meanings:
1. Smallest maximum distance to the other points. 2. Smallest average distance to other points. 3. Smallest sum of squares of distances to other points. 4. Etc., etc.
clustroid 1 3 2 6 5 intercluster distance
8 4 clustroid Other Approaches to Defining "Nearness" of Clusters
x Approach 2: intercluster distance = minimum of the distances between any two points, one from each cluster. x Approach 3: Pick a notion of "cohesion" of clusters, e.g., maximum distance from the clustroid. Merge clusters whose union is most cohesive. 9 Cohesion
x Approach 1: Use the diameter of the merged cluster = maximum distance between points in the cluster. x Approach 2: Use the average distance between points in the cluster. 10 Cohesion (2)
x Approach 3: Use a densitybased approach: take the diameter or average distance, e.g., and divide by the number of points in the cluster. Perhaps raise the number of points to a power first, e.g., squareroot. 11 k Means Algorithm(s)
x Assumes Euclidean space. x Start by picking k, the number of clusters. x Initialize clusters by picking one point per cluster. Example: pick one point at random, then k 1 other points, each as far away as possible from the previous points.
12 Populating Clusters
1. For each point, place it in the cluster whose current centroid it is nearest. 2. After all points are assigned, fix the centroids of the k clusters. 3. Optional: reassign all points to their closest centroid. Sometimes moves points between clusters. 13 Example: Assigning Clusters
Reassigned points 6 7 5 x 3 1 8 x 2 4 Clusters after first round
14 Getting k Right Try different k, looking at the change in the average distance to centroid, as k increases. x Average falls rapidly until right k, then changes little.
Average distance to centroid Best value of k k 15 Example: Picking k
Too few; many long distances to centroid. x x x x x x x x xx x x x x x x x x xx x x x x x x x xx x x x x x x x x x x x x
16 Example: Picking k
Just right; distances rather short. x x x x x x x x xx x x x x x x x x xx x x x x x x x xx x x x x x x x x x x x x
17 Example: Picking k
Too many; little improvement in average x distance. x x x x xx x x x x x x x xx x x x x x x xx x x x x x x x x x x x x x x x x x
18 BFR Algorithm
x BFR (BradleyFayyadReina) is a variant of k means designed to handle very large (diskresident) data sets. x It assumes that clusters are normally distributed around a centroid in a Euclidean space. Standard deviations in different dimensions may vary.
19 BFR (2)
x Points are read one mainmemoryfull at a time. x Most points from previous memory loads are summarized by simple statistics. x To begin, from the initial load we select the initial k centroids by some sensible approach.
20 Initialization: k Means
x Possibilities include:
1. Take a small random sample and cluster optimally. 2. Take a sample; pick a random point, and then k 1 more points, each as far from the previously selected points as possible. 21 Three Classes of Points
1. The discard set : points close enough to a centroid to be summarized. 1. The compression set : groups of points that are close together but not close to any centroid. They are summarized, but not assigned to a cluster. 2. The retained set : isolated points.
22 Summarizing Sets of Points
x For each cluster, the discard set is summarized by: 1. The number of points, N. 2. The vector SUM, whose i th component is the sum of the coordinates of the points in the i th dimension. 3. The vector SUMSQ: i th component = sum of squares of coordinates in i th dimension.
x 2d + 1 values represent any number of points. x Averages in each dimension (centroid coordinates) can be calculated easily as SUMi /N. SUMi = i th component of SUM. d = number of dimensions. 24 Comments (2)
x Variance of a cluster's discard set in dimension i can be computed by: (SUMSQi /N ) (SUMi /N )2 x And the standard deviation is the square root of that. x The same statistics can represent any compression set.
25 "Galaxies" Picture
Points in the RS Compressed sets. Their points are in the CS. A cluster. Its points are in the DS. The centroid
26 Processing a "MemoryLoad" of Points
1. Find those points that are "sufficiently close" to a cluster centroid; add those points to that cluster and the DS. 2. Use any mainmemory clustering algorithm to cluster the remaining points and the old RS.
x Clusters go to the CS; outlying points to the RS.
27 Processing (2)
1. Adjust statistics of the clusters to account for the new points. 2. Consider merging compressed sets in the CS. 3. If this is the last round, merge all compressed sets in the CS and all RS points into their nearest cluster.
28 A Few Details . . .
x How do we decide if a point is "close enough" to a cluster that we will add the point to that cluster? x How do we decide whether two compressed sets deserve to be combined into one? 29 How Close is Close Enough?
x We need a way to decide whether to put a new point into a cluster. x BFR suggest two ways: 1. The Mahalanobis distance is less than a threshold. 2. Low likelihood of the currently nearest centroid changing.
30 Mahalanobis Distance
x Normalized Euclidean distance. x For point (x1,...,xk) and centroid (c1,...,ck):
1. Normalize in each dimension: yi = (xi ci)/ i 2. Take sum of the squares of the yi 's. 3. Take the square root. 31 Mahalanobis Distance (2)
x If clusters are normally distributed in d dimensions, then after transformation, one standard deviation = d. x Accept a point for a cluster if its M.D. is < some threshold, e.g. 4 standard deviations.
32 I.e., 70% of the points of the cluster will have a Mahalanobis distance < d. Picture: Equal M.D. Regions 2 33 Should Two CS Subclusters Be Combined?
x Compute the variance of the combined subcluster. x Combine if the variance is below some threshold. x Many alternatives: treat dimensions differently, consider density.
34 N, SUM, and SUMSQ allow us to make that calculation quickly. The CURE Algorithm
x Problem with BFR/k means: Assumes clusters are normally distributed in each dimension. And axes are fixed ellipses at an angle are not OK. Assumes a Euclidean distance. Allows clusters to assume any shape.
35 x CURE: Example: Stanford Faculty Salaries
h e h h h e h e h h e e e e salary e h h h h h age e e e h 36 Starting CURE
1. Pick a random sample of points that fit in main memory. 2. Cluster these points hierarchically group nearest points/clusters. 3. For each cluster, pick a sample of points, as dispersed as possible. 4. From the sample, pick representatives by moving them (say) 20% toward the centroid of the cluster.
37 Example: Initial Clusters
h e h h h e h e h h e e e e salary e h h h h h age
38 e e e h Example: Pick Dispersed Points
h e h h h e h e h h e e Pick (say) 4 remote points for each cluster. e e salary e h h h h h age e e e h 39 Example: Pick Dispersed Points
h e h h h e h e h h e e Move points (say) 20% toward the centroid. e e salary e h h h h h age e e e h 40 Finishing CURE
x Now, visit each point p in the data set. x Place it in the "closest cluster." Normal definition of "closest": that cluster with the closest (to p ) among all the sample points of all the clusters. 41 ...
View Full Document
This document was uploaded on 03/04/2012.
- Fall '09