clustering methods
DESCRIPTION
Clustering Analysis (of Spatial Data and using Peano Count Trees) (Ptree technology is patented by NDSU) Notes: 1. over 100 slides – not going to go through each in detail. Clustering Methods. A Categorization of Major Clustering Methods Partitioning methods Hierarchical methods - PowerPoint PPT PresentationTRANSCRIPT
Clustering Analysis(of Spatial Data and using Peano Count Trees) (Ptree technology is patented by NDSU)
Notes:1. over 100 slides – not going to go through each in detail.
Clustering Methods
A Categorization of Major Clustering MethodsPartitioning methodsHierarchical methodsDensity-based methodsGrid-based methodsModel-based methods
Clustering Methods based on Partitioning
Partitioning method: Construct a partition of a database D of n objects into a set of k clusters
Given a k, find a partition of k clusters that optimizes the chosen partitioning criterion– k-means (MacQueen’67): Each cluster is represented by the
center of the cluster– k-medoids or PAM method (Partition Around Medoids)
(Kaufman & Rousseeuw’87): Each cluster is represented by 1 object in the cluster (~ the middle object or median-like object)
The K-Means Clustering MethodGiven k, the k-means algorithm is implemented in 4 steps (assumes partitioning
criteria is: maximize intra-cluster similarity and minimize inter-cluster similarity. Of course, a heuristic is used. Method isn’t really an optimization)
1. Partition objects into k nonempty subsets (or pick k initial means).
2. Compute the mean (center) or centroid of each cluster of the current partition (if one started with k means, then this step is done).
centroid ~ point that minimizes the sum of dissimilarities from the mean or the sum of the square errors from the mean.
Assign each object to the cluster with the most similar (closest) center.
3. Go back to Step 2
4. Stop when the new set of means doesn’t change (or some other stopping condition?)
k-Means
0
1
2
3
4
5
6
7
8
9
10
0 1 2 3 4 5 6 7 8 9 100
1
2
3
4
5
6
7
8
9
10
0 1 2 3 4 5 6 7 8 9 10
0
1
2
3
4
5
6
7
8
9
10
0 1 2 3 4 5 6 7 8 9 100
1
2
3
4
5
6
7
8
9
10
0 1 2 3 4 5 6 7 8 9 10
Step 1
Step 2
Step 3
Step 4
The K-Means Clustering Method Strength
– Relatively efficient: O(tkn),• n is # objects,• k is # clusters• t is # iterations. Normally, k, t << n.
Weakness– Applicable only when mean is defined (e.g., a vector space)
– Need to specify k, the number of clusters, in advance.
– It is sensitive to noisy data and outliers since a small number of such data can substantially influence the mean value.
The K-Medoids Clustering Method Find representative objects, called medoids, (must be an actual object in the
cluster, where as the mean seldom is).
PAM (Partitioning Around Medoids, 1987)– starts from an initial set of medoids– iteratively replaces one of the medoids by a non-medoid– if it improves the aggregate similarity measure, retain the swap. Do this over all
medoid-nonmedoid pairs– PAM works for small data sets. Does not scale for large data sets
CLARA (Clustering LARge Applications) (Kaufmann,Rousseeuw, 1990) Sub-samples then apply PAM
CLARANS (Clustering Large Applications based on RANdomSearch) (Ng & Han, 1994): Randomized the sampling
PAM (Partitioning Around Medoids) (1987)
Use real object to represent the cluster– Select k representative objects arbitrarily
– For each pair of non-selected object h and selected object i, calculate the total swapping cost TCi,h
– For each pair of i and h,
• If TCi,h < 0, i is replaced by h
• Then assign each non-selected object to the most similar representative object
– repeat steps 2-3 until there is no change
CLARA (Clustering Large Applications) (1990)
CLARA (Kaufmann and Rousseeuw in 1990) It draws multiple samples of the data set, applies PAM on
each sample, and gives the best clustering as the output Strength: deals with larger data sets than PAM Weakness:
– Efficiency depends on the sample size
– A good clustering based on samples will not necessarily represent a good clustering of the whole data set if the sample is biased
CLARANS (“Randomized” CLARA) (1994)
CLARANS (A Clustering Algorithm based on Randomized Search) (Ng and Han’94)
CLARANS draws sample of neighbors dynamically The clustering process can be presented as searching a graph
where every node is a potential solution, that is, a set of k medoids If the local optimum is found, CLARANS starts with new
randomly selected node in search for a new local optimum (Genetic-Algorithm-like)
Finally the best local optimum is chosen after some stopping condition.
It is more efficient and scalable than both PAM and CLARA
Distance-based partitioning has drawbacks
Simple and fast O(N) The number of clusters, K, has to be arbitrarily chosen before it is known how many
clusters is correct. Produces round shaped clusters, not arbitrary shapes (Chameleon data set below)
Sensitive to the selection of the initial partition and may converge to a local minimum of the criterion function if the initial partition is not well chosen.
Correct result
K-means result
Distance-based partitioning (Cont.) If we start with A, B, and C as the
initial centriods around which the three clusters are built, then we end up with the partition {{A}, {B, C}, {D, E, F, G}} shown by ellipses.
Whereas, the correct three-cluster solution is obtained by choosing, for example, A, D, and F as the initial cluster means (rectangular clusters).
A Vertical Data Approach Partition the data set using rectangle P-trees (a gridding)
These P-trees can be viewed as a grouping (partition) of data
Pruning out outliers by disregard those sparse values
Input: total number of objects (N), percentage of outliers (t) Output: Grid P-trees after prune
(1) Choose the Grid P-tree with smallest root count (Pgc)(2) outliers:=outliers OR Pgc
(3) if (outliers/N<= t) then remove Pgc and repeat (1)(2)
Finding clusters using PAM method; each grid P-tree is an object– Note: when we have a P-tree mask for each cluster, the mean is just
the vector sum of the basic Ptrees ANDed with the cluster Ptree,divided by the rootcount of the cluster Ptree
Distance Function
Data Matrix n objects × p variables
Dissimilarity Matrix n objects × n objects
X11 … X1f … X1p. . .. . .. . .
Xi1 … Xif … Xip. . .. . .. . .
Xn1 … Xnf … Xnp
0 d(2,1) 0 d(3,1) d(3,2) 0 . . . . . . . . . . . . . . .
d(n,1) d(n,2) … d(n,n-1) 0
2222
211 )()()(j)d(i,
:j and iobject between DistanceEuclidian
jpipjiji XXXXXX
AGNES (Agglomerative Nesting) Introduced in Kaufmann and Rousseeuw (1990) Use the Single-Link (distance between two sets is the
minimum pairwise distance) method Merge nodes that are most similarity Eventually all nodes belong to the same cluster
0
1
2
3
4
5
6
7
8
9
10
0 1 2 3 4 5 6 7 8 9 100
1
2
3
4
5
6
7
8
9
10
0 1 2 3 4 5 6 7 8 9 100
1
2
3
4
5
6
7
8
9
10
0 1 2 3 4 5 6 7 8 9 10
DIANA (Divisive Analysis) Introduced in Kaufmann and Rousseeuw (1990)
Inverse order of AGNES (intitially all objects are in one cluster; then it is split according to some criteria (e.g., maximize some aggregate measure of pairwise dissimilarity again)
Eventually each node forms a cluster on its own
0
1
2
3
4
5
6
7
8
9
10
0 1 2 3 4 5 6 7 8 9 100
1
2
3
4
5
6
7
8
9
10
0 1 2 3 4 5 6 7 8 9 10
0
1
2
3
4
5
6
7
8
9
10
0 1 2 3 4 5 6 7 8 9 10
Contrasting Clustering Techniques
Partitioning algorithms: Partition a dataset to k clusters, e.g., k=3
Hierarchical alg: Create hierarchical decomposition of ever-finer partitions.
e.g., top down (divisively).
bottom up (agglomerative)
Hierarchical Clustering
a b
Step 1
d e
Step 2
c d e
Step 3
a b c d e
Step 4
b
dc
e
a
Step 0 Agglomerative
Hierarchical Clustering (top down)
In either case, one gets a nice dendogram in which any maximal anti-chain (no 2 nodes linked) is a clustering (partition).
ba
Step 4
de
Step 3
c
d e
Step 2
a b
c d e
Step 1
a b c d e
Step 0Divisive
Hierarchical Clustering (Cont.)
Recall that any maximal anti-chain (maximal set of nodes in whichno 2 are chained) is a clustering (a dendogram offers many).
Hierarchical Clustering (Cont.)
But the “horizontal” anti-chains are the clusterings resulting from thetop down (or bottom up) method(s).
Hierarchical Clustering (Cont.) Most hierarchical clustering algorithms are variants of the single-
link, complete-link or average link.
Of these, single-link and complete link are most popular.
– In the single-link method, the distance between two clusters is the minimum of the distances between all pairs of patterns drawn one from each cluster.
– In the complete-link algorithm, the distance between two clusters is the maximum of all pairwise distances between pairs of patterns drawn one from each cluster.
– In the average-link algorithm, the distance between two clusters is the average of all pairwise distances between pairs of patterns drawn one from each cluster (which is the same as the distance between the means in the vector space case – easier to calculate).
Distance Between Clusters
Single Link: smallest distance between any pair of points from two clusters
Complete Link: largest distance between any pair of points from two clusters
Distance between Clusters (Cont.)
Average Link: average distance between points from two clusters
• Centroid: distance between centroids of the two clusters
Single Link vs. Complete Link (Cont.)
Single link works but not complete link Complete link
works but not single link
Complete link works but not
single link
Single Link vs. Complete Link (Cont.)
Single link works
1
1 1
1
1
1
1
2 222
Complete link doesn’t
1
1 1
1
1
1
1
2 222
Single Link vs. Complete Link (Cont.)
Single link doesn’t works
11
2
2
2
11
22
22
Complete link does
11
2
2
2
11
22
22
1-cluster noise 2-cluster
Hierarchical vs. Partitional
Hierarchical algorithms are more versatile than partitional algorithms. – For example, the single-link clustering algorithm works well on data
sets containing non-isotropic (non-roundish) clusters including well-separated, chain-like, and concentric clusters, whereas a typical partitional algorithm such as the k-means algorithm works well only on data sets having isotropic clusters.
On the other hand, the time and space complexities of the partitional algorithms are typically lower than those of the hierarchical algorithms.
More on Hierarchical Clustering Methods Major weakness of agglomerative clustering methods
– do not scale well: time complexity of at least O(n2), where n is the number of total objects
– can never undo what was done previously (greedy algorithm)
Integration of hierarchical with distance-based clustering– BIRCH (1996): uses Clustering Feature tree (CF-tree) and
incrementally adjusts the quality of sub-clusters
– CURE (1998): selects well-scattered points from the cluster and then shrinks them towards the center of the cluster by a specified fraction
– CHAMELEON (1999): hierarchical clustering using dynamic modeling
Density-Based Clustering Methods
Clustering based on density (local cluster criterion), such as density-connected points
Major features:– Discover clusters of arbitrary shape– Handle noise– One scan– Need density parameters as termination condition
Several interesting studies:– DBSCAN: Ester, et al. (KDD’96)– OPTICS: Ankerst, et al (SIGMOD’99).– DENCLUE: Hinneburg & D. Keim (KDD’98)– CLIQUE: Agrawal, et al. (SIGMOD’98)
Density-Based Clustering: Background Two parameters:
: Maximum radius of the neighbourhood
– MinPts: Minimum number of points in an -neighbourhood of that point
N(p): {q belongs to D | dist(p,q) } Directly (density) reachable: A point p is directly
density-reachable from a point q wrt. , MinPts if – 1) p belongs to N(q)
– 2) q is a core point:
|N(q)| MinPts p
qMinPts = 5
= 1 cm
Density-Based Clustering: Background (II)
Density-reachable: – A point p is density-reachable from a point q (p)
wrt , MinPts if there is a chain of points p1, …, pn, p1=q, pn=p such that pi+1 is directly density-reachable from pi
q, q is density-reachable from q.• Density reachability is reflexive and transitive, but
not symmetric, since only core objects can be density reachable to each other.
Density-connected– A point p is density-connected to a q wrt , MinPts
if there is a point o such that both, p and q are density-reachable from o wrt , MinPts.
– Density reachability is not symmetric, Density connectivity inherits the reflexivity and transitivity and provides the symmetry. Thus, density connectivity is an equivalence relation and therefore gives a partition (clustering).
p
qp1
p q
o
DBSCAN: Density Based Spatial Clustering of Applications with Noise
Relies on a density-based notion of cluster: A cluster is defined as an equivalence class of density-connected points.– Which gives the transitive property for the density connectivity binary relation
and therefore it is an equivalence relation whose components form a partition (clustering) according to the duality.
Discovers clusters of arbitrary shape in spatial databases with noise
Core
Border
Outlier
= 1cm
MinPts = 3
DBSCAN: The Algorithm– Arbitrary select a point p
– Retrieve all points density-reachable from p wrt , MinPts.
– If p is a core point, a cluster is formed (note: it doesn’t matter which of the core points within a cluster you start at since density reachability is symmetric on core points.)
– If p is a border point or an outlier, no points are density-reachable from p and DBSCAN visits the next point of the database. Keep track of such points. If they don’t get “scooped up” by a later core point, then they are outliers.
– Continue the process until all of the points have been processed.
– What about a simpler version of DBSCAN:• Define core points and core neighborhoods the same way.
• Define (undirected graph) edge between two points if they cohabitate a core nbrhd.
• The connectivity component partition is the clustering.
– Other related method? How does vertical technology help here? Gridding?
OPTICS
Ordering Points To Identify Clustering Structure– Ankerst, Breunig, Kriegel, and Sander (SIGMOD’99)– http://portal.acm.org/citation.cfm?id=304187
– Addresses the shortcoming of DBSCAN, namely choosing parameters.
– Develops a special order of the database wrt its density-based clustering structure
– This cluster-ordering contains info equivalent to the density-based clusterings corresponding to a broad range of parameter settings
– Good for both automatic and interactive cluster analysis, including finding intrinsic clustering structure
OPTICS
Does this order resemble theTotal Variation order?
Reachability-distance
Cluster-orderof the objects
undefined
‘
DENCLUE: using density functions
DENsity-based CLUstEring by Hinneburg & Keim (KDD’98) Major features
– Solid mathematical foundation– Good for data sets with large amounts of noise– Allows a compact mathematical description of arbitrarily
shaped clusters in high-dimensional data sets– Significant faster than existing algorithm (faster than
DBSCAN by a factor of up to 45 – claimed by authors ???)– But needs a large number of parameters
Uses grid cells but only keeps information about grid cells that do actually contain data points and manages these cells in a tree-based access structure.
Influence function: describes the impact of a data point within its neighborhood.– F(x,y) measures the influence that y has on x.– A very good influence function is the Gaussian, F(x,y) = e –d2(x,y)/2
– Others include functions similar to the squashing functions used in neural networks.– One can think of the influence function as a measure of the contribution to the
density at x made by y. Overall density of the data space can be calculated as the sum of the influence function
of all data points. Clusters can be determined mathematically by identifying density attractors. Density attractors are local maximal of the overall density function.
Denclue: Technical Essence
DENCLUE(D,σ,ξc,ξ)
1. Grid Data Set (use r = σ, the std. dev.)2. Find (Highly) Populated Cells (use a
threshold=ξc) (shown in blue)3. Identify populated cells (+nonempty cells)
4. Find Density Attractor pts, C*, using hill climbing:
1. Randomly pick a point, pi.2. Compute local density (use r=4σ)3. Pick another point, pi+1, close to pi,
compute local density at pi+1
4. If LocDen(pi) < LocDen(pi+1), climb5. Put all points within distance σ/2 of path,
pi, pi+1, …C* into a “density attractor cluster called C*
5. Connect the density attractor clusters, using a threshold, ξ, on the local densities of the attractors.
A. Hinneburg and D. A. Keim. An Efficient Approach to Clustering in Multimedia Databases with Noise. In Proc. 4th Int. Conf. on Knowledge Discovery and Data Mining. AAAI Press, 1998. & KDD 99 Workshop.
Comparison: DENCLUE Vs DBSCAN
BIRCH (1996) Birch: Balanced Iterative Reducing and Clustering using Hierarchies, by
Zhang, Ramakrishnan, Livny (SIGMOD’96http://portal.acm.org/citation.cfm?id=235968.233324&dl=GUIDE&dl=ACM&idx=235968&part=periodical&WantType=periodical&title
=ACM%20SIGMOD%20Record&CFID=16013608&CFTOKEN=14462336
Incrementally construct a CF (Clustering Feature) tree, a hierarchical data structure for multiphase clustering– Phase 1: scan DB to build an initial in-memory CF tree (a multi-level compression
of the data that tries to preserve the inherent clustering structure of the data) – Phase 2: use an arbitrary clustering algorithm to cluster the leaf nodes of the CF-tree
Scales linearly: finds a good clustering with a single scan and improves quality with a few additional scans
Weakness: handles only numeric data, and sensitive to the order of the data record.
ABSTRACT
Finding useful patterns in large datasets has attracted considerable interest recently, and one of the most widely studied problems in this area is the identification of clusters, or densely populated regions, in a multi-dimensional dataset.
Prior work does not adequately address the problem of large datasets and minimization of I/O costs.
This paper presents a data clustering method named BIRCH (Balanced Iterative Reducing and Clustering using Hierarchies), and demonstrates that it is especially suitable for very large databases.
BIRCH incrementally and dynamically clusters incoming multi-dimensional metric data points to try to produce the best quality clustering with the available resources (i.e., available memory and time constraints).
BIRCH can typically find a good clustering with a single scan of the data, and improve the quality further with a few additional scans.
BIRCH is also the first clustering algorithm proposed in the database area to handle "noise" (data points that are not part of the underlying pattern) effectively.
We evaluate BIRCH's time/space efficiency, data input order sensitivity, and clustering quality through several experiments.
BIRCH
Clustering Feature VectorClustering Feature: CF = (N, LS, SS)
N: Number of data points
LS: Ni=1=Xi
SS: Ni=1=Xi
2
Branching factor = max # children
Threshold = max diameter of leaf cluster
0
1
2
3
4
5
6
7
8
9
10
0 1 2 3 4 5 6 7 8 9 10
CF = (5, (16,30),(54,190))
(3,4)(2,6)(4,5)(4,7)(3,8)
Birch
Branching factor, B=6
Threshold, L = 7CF1
child1
CF3
child3
CF2
child2
CF6
child6
CF1
child1
CF3
child3
CF2
child2
CF5
child5
CF1 CF2 CF6prev next CF1 CF2 CF4
prev next
Root
Non-leaf node
Leaf node Leaf node
Iteratively put points into closest leaf until threshold is exceed, then split leaf.Inodes summarize their subtrees and Inodes get split when threshold is exceeded.Once in-memory CF tree is built, use another method to cluster leaves together.
CURE (Clustering Using REpresentatives )
CURE: proposed by Guha, Rastogi & Shim, 1998 http://portal.acm.org/citation.cfm?id=276312
– Stops the creation of a cluster hierarchy if a level consists of k clusters
– Uses multiple representative points to evaluate the distance between clusters
– adjusts well to arbitrary shaped clusters (not necessarily distance-based
– avoids single-link effect
Drawbacks of Distance-Based Method
Drawbacks of square-error based clustering method – Consider only one point as representative of a cluster– Good only for convex shaped, similar size and density,
and if k can be reasonably estimated
Cure: The Algorithm
– Very much a hybrid method (involves pieces from many others).
– Draw random sample s.
– Partition sample to p partitions with size s/p
– Partially cluster partitions into s/pq clusters
– Eliminate outliers
• By random sampling
• If a cluster grows too slow, eliminate it.
– Cluster partial clusters.
– Label data in disk
ABSTRACT
Clustering, in data mining, is useful for discovering groups and identifying interesting distributions in the underlying data. Traditional clustering algorithms either favor clusters with spherical shapes and similar sizes, or are very fragile in the presence of outliers.
We propose a new clustering algorithm called CURE that is more robust to outliers, and identifies clusters having non-spherical shapes and wide variances in size.
CURE achieves this by representing each cluster by a certain fixed number of points that are generated by selecting well scattered points from the cluster and then shrinking them toward the center of the cluster by a specified fraction.
Having more than one representative point per cluster allows CURE to adjust well to the geometry of non-spherical shapes and the shrinking helps to dampen the effects of outliers.
To handle large databases, CURE employs a combination of random sampling and partitioning. A random sample drawn from the data set is first partitioned and each partition is partially clustered. The partial clusters are then clustered in a second pass to yield the desired clusters.
Our experimental results confirm that the quality of clusters produced by CURE is much better than those found by existing algorithms.
Furthermore, they demonstrate that random sampling and partitioning enable CURE to not only outperform existing algorithms but also to scale well for large databases without sacrificing clustering quality.
Cure
Data Partitioning and Clustering
– s = 50– p = 2– s/p = 25
x x
x
y
y y
y
x
y
x
s/pq = 5
Cure: Shrinking Representative Points
Shrink the multiple representative points towards the gravity center by a fraction of .
Multiple representatives capture the shape of the cluster
x
y
x
y
Clustering Categorical Data: ROCKhttp://portal.acm.org/citation.cfm?id=351745
ROCK: Robust Clustering using linKs,by S. Guha, R. Rastogi, K. Shim (ICDE’99). – Agglomerative Hierarchical– Use links to measure similarity/proximity– Not distance based– Computational complexity:
Basic ideas:– Similarity function and neighbors:
Let T1 = {1,2,3}, T2={3,4,5}
O n nm m n nm a( log )2 2
Sim T TT TT T
( , )1 21 2
1 2
Sim T T( , ){ }
{ , , , , }.1 2
31 2 3 4 5
15
0 2
ROCKAbstractClustering, in data mining, is useful to discover distribution patterns in the underlying
data.Clustering algorithms usually employ a distance metric based (e.g., euclidean)
similarity measure in order to partition the database such that data points in the same partition are more similar than points in different partitions.
In this paper, we study clustering algorithms for data with boolean and categorical attributes.
We show that traditional clustering algorithms that use distances between points for clustering are not appropriate for boolean and categorical attributes. Instead, we propose a novel concept of links to measure the similarity/proximity between a pair of data points.
We develop a robust hierarchical clustering algorithm ROCK that employs links and not distances when merging clusters.
Our methods naturally extend to non-metric similarity measures that are relevant in situations where a domain expert/similarity table is the only source of knowledge.
In addition to presenting detailed complexity results for ROCK, we also conduct an experimental study with real-life as well as synthetic data sets to demonstrate the effectiveness of our techniques.
For data with categorical attributes, our findings indicate that ROCK not only generates better quality clusters than traditional algorithms, but it also exhibits good scalability properties.
Rock: Algorithm
Links: The number of common neighbors for the two pts
Algorithm– Draw random sample– Cluster with links– Label data in disk
{1,2,3}, {1,2,4}, {1,2,5}, {1,3,4}, {1,3,5}{1,4,5}, {2,3,4}, {2,3,5}, {2,4,5}, {3,4,5}
{1,2,3} {1,2,4}3
CHAMELEON
CHAMELEON: hierarchical clustering using dynamic modeling, by G. Karypis, E.H. Han and V. Kumar’99
http://portal.acm.org/citation.cfm?id=621303
Measures the similarity based on a dynamic model– Two clusters are merged only if the interconnectivity and
closeness (proximity) between two clusters are high relative to the internal interconnectivity of the clusters and closeness of items within the clusters
A two phase algorithm– 1. Use a graph partitioning algorithm: cluster objects into a large
number of relatively small sub-clusters– 2. Use an agglomerative hierarchical clustering algorithm: find the
genuine clusters by repeatedly combining these sub-clusters
CHAMELEONABSTRACT
Many advanced algorithms have difficulty dealing with highly variable clusters that do not follow a preconceived model.
By basing its selections on both interconnectivity and closeness, the Chameleon algorithm yields accurate results for these highly variable clusters.
Existing algorithms use a static model of the clusters and do not use information about the nature of individual clusters as they are merged.
Furthermore, one set of schemes (the CURE algorithm and related schemes) ignores the information about the aggregate interconnectivity of items in two clusters.
Another set of schemes (the Rock algorithm, group averaging method, and related schemes) ignores information about the closeness of two clusters as defined by the similarity of the closest items across two clusters.
By considering either interconnectivity or closeness only, these algorithms can select and merge the wrong pair of clusters.
Chameleon's key feature is that it accounts for both interconnectivity and closeness in identifying the most similar pair of clusters.
Chameleon finds the clusters in the data set by using a two-phase algorithm.During the first phase, Chameleon uses a graph-partitioning algorithm to cluster the
data items into several relatively small subclusters.During the second phase, it uses an algorithm to find the genuine clusters by repeatedly
combining these sub-clusters.
Overall Framework of CHAMELEON
Construct
Sparse Graph Partition the Graph
Merge Partition
Final Clusters
Data Set
Grid-Based Clustering Method
Using multi-resolution grid data structure Several interesting methods
– STING (a STatistical INformation Grid approach) by Wang, Yang and Muntz (1997)
– WaveCluster by Sheikholeslami, Chatterjee, and Zhang (VLDB’98)
• A multi-resolution clustering approach using wavelet method
– CLIQUE: Agrawal, et al. (SIGMOD’98)
Vertical griddingWe can observe that almost all methods discussed so far suffer from the curse of cardinality (for very large cardinality data sets, the algorithms are too slow to finish in the average life time!) and/or the curse of dimensionality (points are all at ~ same distance).
The work-arounds employed to address the curses
sampling (throw out most of the points in a way that what remains is low enough cardinality for the algorithm to finish and in such a way that the remaining “sample” contains all the information of the original data set (Therein is the problem – that is impossible to do in general);
Gridding (agglomerate all points in a grid cell and treat them as one point (smooth the data set to this gridding level). The problem with gridding, often, is that info is lost and the data structure that holds the grid cell information is very complex. With vertical methods (e.g., P-trees), all the info can be retained and griddings can be constructed very efficiently on demand. Horizontal data structures can’t do this.
Subspace restrictions (e.g., Principal Components, Subspace Clustering…)
Gradient based methods (e.g., the gradient tangent vector field of a response surface reduces the calculations to the number of dimensions, not the number of combinations of dimensions.)
j-hi gridding: the j hi order bits identify a grid cells and the rest identify points in a particular cell. Thus, j-hi cells are not necessarily cubical (unless all attribute bit-widths are the same).
j-lo gridding; the j lo order bits identify points in a particular cell and the rest identify a grid cell. Thus, j-lo cells always have a nice uniform shape (cubical).
1-hi gridding of Vector Space, R(A1, A2, A3) in which all bit-widths are the same = 3 (so each grid cell contains 22 * 22 * 22 = 64 potential points). Grid cells are identified by their Peano id (Pid) internally the point’s cell coordinates are shown - called the grid cell id and cell points are id’ed by coordinates within their cell.
A1
hi-bit
A3
hi-bit
A2
hi-bit
0
1
0 1
0
1
gci=001 gcp=
00,00,00
Pid = 001
gci=001 gcp=
00,00,01
gci=001 gcp=
00,01,00 gci=001
gcp=00,01,01
gci=001 gcp=
01,00,00 gci=001
gcp=01,00,01
gci=001 gcp=
01,01,00 gci=001
gcp=01,01,01
gci=001 gcp=
00,00,10 gci=001 gcp=
00,00,11
gci=001 gcp=
00,01,10 gci=001 gcp=
00,01,11 gci=001
gcp=01,00,10 gci=001
gcp=01,00,11
gci=001 gcp=
01,01,10 gci=001 gcp=
01,01,11
gci=001 gcp=
00,10,00 gci=001 gcp=
00,10,01
gci=001 gcp=
00,11,00 gci=001 gcp=
00,11,01 gci=001
gcp=01,10,00 gci=001
gcp=01,10,01
gci=001 gcp=
01,11,00 gci=001 gcp=
01,11,01
gci=001 gcp=
00,10,10 gci=001 gcp=
00,10,11
gci=001 gcp=
00,11,10 gci=001 gcp=
00,11,11 gci=001
gcp=01,10,10
gci=001 gcp=
01,10,11
gci=001 gcp=
01,11,10 gci=001 gcp=
01,11,11
gci=001 gcp=
10,00,00 gci=001
gcp=10,00,01
gci=001 gcp=
10,01,00 gci=001
gcp=10,01,01
gci=001 gcp=
10,00,10 gci=001
gcp=10,00,11
gci=001 gcp=
10,01,10 gci=001
gcp=10,01,11
gci=001 gcp=
10,10,00 gci=001 gcp=
10,10,01
gci=001 gcp=
10,11,00 gci=001 gcp=
10,11,01
gci=001 gcp=
10,10,10 gci=001
gcp=10,10,11
gci=001 gcp=
10,11,10 gci=001 gcp=
10,11,11
gci=001 gcp=
11,00,00 gci=001 gcp=
11,00,01
gci=001 gcp=
11,01.00 gci=001 gcp=
11,01.01
gci=001 gcp=
11,00,10 gci=001
gcp=11,00,11
gci=001 gcp=
11,01,10 gci=001
gcp=11,01,11
gci=001 gcp=
11,10,00 gci=001 gcp=
11,10,01
gci=001 gcp=
11,11,00 gci=001
gcp=11,11,01
gci=001 gcp=
11,10,10 gci=001 gcp=
11,10,11
gci=001 gcp=
11,11,10 gci=001 gcp=
11,11,11
A1
A3
A2
00 01 10 11
00
01
10
11
00
0110
11
Pid = 001.001
gci=00,00,11
gcp=0,0,0 gci=00,00,11
gcp=0,0,1
gci=00,00,11
gcp=0,1,0 gci=00,00,11
gcp=0,1,1 gci=00,00,11
gcp=1,0,0 gci=00,00,11
gcp=1,0,1
gci=00,00,11
gcp=1,1,0 gci=00,00,11
gcp=1,1,1
2-hi gridding of Vector Space, R(A1, A2, A3) in which all bitwidths are the same = 3 (so each grid cell contains 21 * 21 * 21 = 8 points).
1-hi gridding of R(A1, A2, A3), bitwidths of 3,2,3
A1
hi-bit
A3
hi-bit
A2
hi-bit
0
1
0 1
0
1
gci=001 gcp=
00,0,00
Pid = 001
gci=001 gcp=
00,0,01
gci=001 gcp=
00,1,00 gci=001
gcp=00,1,01
gci=001 gcp=
01,0,00 gci=001 gcp=
01,0,01
gci=001 gcp=
01,1,00 gci=001
gcp=01,1,01
gci=001 gcp=
00,0,10 gci=001 gcp=
00,0,11
gci=001 gcp=
00,1,10 gci=001 gcp=
00,1,11 gci=001
gcp=01,0,10
gci=001 gcp=
01,0,11
gci=001 gcp=
01,1,10 gci=001 gcp=
01,1,11
gci=001 gcp=
10,0,00 gci=001 gcp=
10,0,01
gci=001 gcp=
10,1,00 gci=001 gcp=
10,1,01
gci=001 gcp=
10,0,10 gci=001
gcp=10,0,11
gci=001 gcp=
10,1,10 gci=001
gcp=10,1,11
gci=001 gcp=
11,0,00 gci=001 gcp=
11,0,01
gci=001 gcp=
11,1,00 gci=001 gcp=
11,1.01
gci=001 gcp=
11,0,10 gci=001
gcp=11,0,11
gci=001 gcp=
11,1,10 gci=001
gcp=11,1,11
2-hi gridding) of R(A1, A2, A3), bitwidths of 3,2,3 (each grid cell contains 21 * 20 * 21 = 4 potential pts).
A1
2-hi-bit
A3
2-hi-bit
A2
2-hi-bit
00
01
00 01
1011
gcp=0,,0gcp=
0,,1
gcp=1,,0gcp=
1,,1
10 11
0001
10
11
Pid = 3.1.3
gcp=1010,1010,1011
gcp=1010,1010,1010
HOBBit disks and rings: (HOBBit = Hi Order Bifurcation Bit) 4-lo grid where A1,A2,A3 have bit-widths, b1+1, b2+1, b3+1, HOBBit grid centers are points of the form (exactly one per grid cell):x=(x1,b1
..x1,41010, x2,b2..x2,41010, x3,b3
..x3,41010) where xi,js range over all binary patterns
HOBBit disk about x, of radius 20 , H(x,20).
Note: we have switched the direction of A3
A1
A3
A2(x1,b1
..x1,41010, x2,b2..x2,41010, x3,b3
..x3,41010)
(x1,b1..x1,41010, x2,b2
..x2,41011, x3,b3..x3,41010)
(x1,b1..x1,41011, x2,b2
..x2,41010, x3,b3..x3,41011)(x1,b1
..x1,41010, x2,b2..x2,41010, x3,b3
..x3,41011)
(x1,b1..x1,41010, x2,b2
..x2,41011, x3,b3..x3,41011)
(x1,b1..x1,41011, x2,b2
..x2,41010, x3,b3..x3,41010)
gcp=1010,1011,1011
gcp=1010,1011,1010 gcp=
1011,1010,1011
gcp=1011,1010,1010
gcp=1011,1011,1011
gcp=1011,1011,1010
(x1,b1..x1,41011, x2,b2
..x2,41011, x3,b3..x3,41011)
(x1,b1..x1,41011, x2,b2
..x2,41011, x3,b3..x3,41010)
H(x,21) HOBBit disk about a HOBBit grid center pt, x = , of radius 21
A1
A2 A3
(x1,b1..x1,41000, x2,b2
..x2,41000, x3,b3..x3,41000)
(x1,b1..x1,41011, x2,b2
..x2,41011, x3,b3..x3,41011)
(x1,b1..x1,41000, x2,b2
..x2,41011, x3,b3..x3,41000)
(x1,b1..x1,41010, x2,b2
..x2,41010, x3,b3..x3,41010)
(x1,b1..x1,41011, x2,b2
..x2,41000, x3,b3..x3,41011)
(x1,b1..x1,41011, x2,b2
..x2,41000, x3,b3..x3,41000)
(x1,b1..x1,41011, x2,b2
..x2,41000, x3,b3..x3,41010)
(x1,b1..x1,41011, x2,b2
..x2,41000, x3,b3..x3,41001)
(x1,b1..x1,41011, x2,b2
..x2,41010, x3,b3..x3,41011)
(x1,b1..x1,41000, x2,b2
..x2,41011, x3,b3..x3,41011)
(x1,b1..x1,41010, x2,b2
..x2,41010, x3,b3..x3,41010)
The Regions of H(x,21) are as follows:
A1
A2
A3
These REGIONS are labeled with dimensions in which length is increased(e.g., all three dimensions are increased below).
A1
A2
A3
123-REGION
A1
A2
A3
13-REGION
A1
A2
A3
23-REG
A1
A2
A3
12-REGION
A1
A2
A3
3-REG
A1
A2
A3
2-REG
A1
A2
A3
1-REGION
A1
A2
A3
H(x,20) =123-REGOf H(x,20)
1-REGION
A1
A2 A3
H(x,20
2-REG
3-REG
12-REGION
23-REG
13-REGION
123-REGION
Algorithm (for computing gradients):
1. Select an “outlier threshold”, (pts without neighbors in their ot L-disk are outliers – That is, there is no gradient at these outlier points (instantaneous rate of response change is zero).
2. Create an j-lo grid with j=ot (see previous slides - where HOBBit disks are built out from HOBBit centers x = ( x1,b1
…x1,ot+11010 , … , xn,bn…xn,ot+11010 ), xi,j
s ranging over all
binary patterns).
3. Pick a point, x in R. Build out alternating one-sided-rings centered at x until a neighbor is found or radius ot is exceeded (in which case x is declared an outlier). If a neighbor is found at a raduis, ri < ot 2j, f/ xk(x) is estimated as below:
1. Note: one can use L-HOBBit or L ordinary distance.2. Note: One-sided means that each successive build out increases aternatively only in the positive direction in all
dimensions then only in the negative direction in all dimensions.3. Note: Building out HOBBit disks from a HOBBit center automatically gives “one-sided” rings (a built-out ring
is defined to be the built-out disk minus the previous built-out disk) as shown in the next few slides.
( RootcountD(x,ri) - RootcountD(x,ri)k ) / xk where D(x,ri)k is D(x,ri-1) expandedin all dimensions except k.
Alternatively in 3., actually calculate the mean (or median?) of the new points encountered in D(x,ri) (we have a P-tree mask for the set so this it trivial) and measure the xk-distance.
NOTE: Might want to go 1 more ring out to see if one gets the same or a similar gradient(this seems particularly important when j is odd (since the gradient then points the opposite way.)NOTE: the calculation of xk can be done in various ways. Which is best?
H(x,21) H(x,21)1
HOBBit center, x=(x1,b1..x1,41010, x2,b2
..x2,41010, x3,b3..x3,41010)First new point
grad
ient
A1
A2 A3
Est f/ xk(x) = (RcD(x,ri) - RootcountD(x,ri)1 ) / x1 =(2-1)/(-1) = -1
H(x,21)
H(x,21)2
grad
ient
( RootcountD(x,ri) - RootcountD(x,ri)2 ) / x2 = (2-1)/(-1) = -1
A1
A2 A3
H(x,21)
H(x,21)3
grad
ient
( RootcountD(x,ri) - RootcountD(x,ri)3 ) / x3 = (2-1)/(-1) = -1
A1
A2 A3
H(x,21)
HOBBit center, x=(x1,b1..x1,41010, x2,b2
..x2,41010, x3,b3..x3,41010)First new point
grad
ient
A1
A2 A3
Est f/ xk(x) = (RcD(x,ri) - RootcountD(x,ri)1 ) / x1 =(2-1)/(-1) = -1
Est f/ xk(x)= ( RcD(x,ri) - RootcountD(x,ri)2 ) / x2 = (2-1)/(-1) = -1
Est f/ xk(x)=( RcD(x,ri) - RootcountD(x,ri)3 ) / x3 = (2-1)/(-1) = -1
Estimated gradient Check other regions too?
H(x,21) H(x,21)1
HOBBit center, x=(x1,b1..x1,41010, x2,b2
..x2,41010, x3,b3..x3,41010)
grad
ient
A1
A2 A3
Est f/ xk(x) = (RcD(x,ri) - RootcountD(x,ri)1 ) / x1 =(2-2)/(-1) = 0
First new point
H(x,21)
H(x,21)2
A1
A2 A3
grad
ient
( RootcountD(x,ri) - RootcountD(x,ri)2 ) / x2 = (2-1)/(-1) = -1
H(x,21)
H(x,21)3
A1
A2 A3
( RootcountD(x,ri) - RootcountD(x,ri)3 ) / x3 = (2-1)/(-1) = -1
grad
ient
H(x,21)A1
A2 A3
grad
ient
( RootcountD(x,ri) - RootcountD(x,ri)2 ) / x2 = (2-1)/(-1) = -1
( RootcountD(x,ri) - RootcountD(x,ri)3 ) / x3 = (2-1)/(-1) = -1
Estimated gradient Check other regions too?
H(x,21) H(x,21)1
HOBBit center, x=(x1,b1..x1,41010, x2,b2
..x2,41010, x3,b3..x3,41010)First new point
A1
A2 A3
Est f/ xk(x) = (RcD(x,ri) - RootcountD(x,ri)1 ) / x1 =(2-1)/(-1) = -1
H(x,21)
H(x,21)2
A1
A2 A3
( RootcountD(x,ri) - RootcountD(x,ri)2 ) / x2 = (2-2)/(-1) = 0
H(x,21)
H(x,21)3
A1
A2 A3
( RootcountD(x,ri) - RootcountD(x,ri)3 ) / x3 = (2-2)/(-1) = 0
H(x,21)
A1
A2 A3
Estimated gradient Check other regions too?Est f/ xk(x) = (RcD(x,ri) - RootcountD(x,ri)1 ) / x1 =(2-1)/(-1) = -1
Intuitively, this Gradient estimation method seems to work.
Next we consider a potential accuracy improvement in which we take the medoid of all new points as the gradient
(or, more accurately, as the point to which we climb in any response surface hill climbing technique)
H(x,21)
new points = s
A1
A2 A3
Estimate the gradient arrowhead as being at the medoid of the new point set (or, more correctly, estimate the next hill-climb step).
Note: If the original points are truly part of a strong cluster, the hill climb will be excellent.
new points centroid =
H(x,21)
new points = s
A1
A2 A3
Estimate the gradient arrowhead as being at the medoid of the new point set (or, more correctly, estimate the next hill-climb step).
Note: If the original points are not truly part of a strong cluster, the weak hill climb will indicate that.
new points centroid =
H(x,22)H(x,22)1
First new point
Est f/ xk(x) = (RcD(x,ri) - RootcountD(x,ri)1 ) / x1 =(2-1)/(3) = 1/3 Est f/ xk(x)= ( RcD(x,ri) - RootcountD(x,ri)2 ) / x2 = (2-1)/(3) = 1/3
Est f/ xk(x)=( RcD(x,ri) - RootcountD(x,ri)3 ) / x3 = (2-1)/(3) = 1/3
A1
A2 A3
Note that the gradient points in the right direction and is very short (as it should be!)
H(x,1)1 H(x,0)
H(x,1)2H(x,1)3
H(x,1)12
H(x,1)23
H(x,1)13
H(x,1)123
H(x,2)
H(x,2)1
H(x,2)1
H(x,2)1
H(x,2)1
H(x,2)1
H(x,2)1
H(x,2)1
To evaluate how well the formula estimates thegradient, it is important to consider all cases of thenew point appearing in one of these regions(if 1 point appears, gradient components areadditive, so it suffices to consider 1?
H(x,1)123
H(x,1)3
H(x,1)2H(x,1)23
H(x,1)1 H(x,0)
H(x,1)12
H(x,1)13
H(x,2)1
H(x,2)2
H(x,2)3
H(x,2)23
H(x,2)13
H(x,2)12
H(x,2)123
To evaluate how well the formula estimates the gradient, it is important to consider all cases of the new point appearing in 1 of these regions (if 1 pt appears, gradient comps add)
H(x,1)123
H(x,1)3
H(x,1)2H(x,1)23
H(x,1)1 H(x,0)
H(x,1)12
H(x,1)13
H(x,2)1
H(x,2)2
H(x,2)3
H(x,2)23
H(x,2)13
H(x,2)12
H(x,2)123
H(x,3)1
H(x,3)13H(x,3)13
H(x,3)123
H(x,3)13
H(x,3)3
H(x,3)13
H( x,23 )Notice that the HOBBit center moves more and more toward the true center as the grid size increases.
Grid based Gradients and Hill Climbing
If we are using gridding to produce the gradient vector field of a response surface, might we always vary xi in the positive direction only? “How can that be done most efficiently?”
1. j-lo gridding, building out HOBBit rings from HOBBit grid centers (see previous slides where this approach was used.) or j-lo gridding. building out HOBBit rings from lo-value grid pts (ending in j 0-bits)
x = ( x1,b1…x1,j+10…0 , … , xn,bn
…xn,j+10…0 )
2. Ordinary j-lo griddng, building out rings from lo-value ids (ending in j zero bits)3. Ordinary j-lo gridding, uilding out Rings from true centers.4. Other? (there are many other possibilities, but we will first explore 2.)
1. 2. Using j-lo gridding with j=3 and lo-value cell identifiers, is shown on the next slide.
3. Of course, we need not use HOBBit build out.
4. With ordinary unit radius build out, the results are more exact, but are the calculations may be more complex???
H(x,1)1H(x,0)
H(x,1)2
H(x,1)3
H(x,1)12
H(x,1)23
H(x,1)13
H(x,1)123
H(x,2)1
H(x,2)1
H(x,2)1
H(x,2)1
H(x,2)1
H(x,2)1
H(x,2)1
H(x,1)1H(x,0)
H(x,1)2
H(x,1)3
H(x,1)12
H(x,1)23
H(x,1)13
H(x,1)123
H(x,2)1
H(x,2)2
H(x,2)3
H(x,2)23
H(x,2)13
H(x,2)12
H(x,2)123
HOBBit j-lo rings using lo-value cell ids x=(x1,b1
…x1,j+10…0 ,…, xn,bn…xn,j+10…0)
Disk(x,0)
Ring(x,1)
Ring(x,2)
= PDisk(x,2)^P’Disk(x,1)
wherePD(x,i) =
Pxb^..^Pxj+1
^P’j^..^P’i+1
= PDisk(x,3)^P’Disk(x,2)
Ordinary j-lo rings using lo-value cell ids x=(x1,b1
…x1,j+10…0 ,…, xn,bn…xn,j+10…0)
k-Medoids Clustering Review: Find representative objects (medoids) (actual objects in the cluster - mean seldom is). PAM (Partitioning Around Medoids)
– Select k representative objects arbitrarily
– For each pair of non-selected object h and selected object i, calculate the total swapping cost TCi,h
– For each pair of i and h,
• If TCi,h < 0, i is replaced by h
• Then assign each non-selected object to the most similar representative object
– repeat steps 2-3 until there is no change
CLARA (Clustering LARge Apps) draws multiple samples of the data set, applies PAM on each sample, and gives the best clustering as the output. Strength: deals with larger data sets than PAM. Weakness: Efficiency depends on the sample size. A good clustering based on samples will not necessarily represent a good clustering of the whole data set if the sample is biased
CLARANS (Clustering Large Apps based on RANdom Search) draws sample of neighbors dynamically. The clustering process can be presented as searching a graph where every node is a potential solution, that is, a set of k medoids. If the local optimum is found, CLARANS starts with new randomly selected node in search for a new local optimum (Genetic-Algorithm-like). Finally the best local optimum is chosen after some stopping condition. It is more efficient and scalable than both PAM and CLARA
A Vertical k-Medoids Clustering Algorithm Following PAM (to illustrate the main killer idea – but it can apply much more widely)
Select k component P-trees
1. The Goal here is to efficiently get one Ptree mask for each component
2. e.g., calculate the smallest j : the j-lo gridding has > k cells.
3. Agglomerate into precisely k components (by ORing Ptrees of cells with closest means/mediods/corners(single_link…)
Where PAM uses: “For each pair of non-selected object h and selected object i, calculate total swapping cost TCi,h. For each pair of i and h, if TCi,h < 0, i is replaced by h, then assign each non-selected object to the most similar object.”, use:
1. Find Medoid of each component, Ci: Calculate TV(Ci,x) x Ci. (create P-tree of all
points with min TV so far. smaller TV, reset this P-tree to it. Ending up with a P-tree, PtMi
of “tieing-Medoids” for Ci. Calc TV(PtMi ,x) x PtMi and pick its medoid (if there are still
multiple, repeat). This is Ci-medoid! Alternatively, just pick 1 pre-Medoid!Note: this avoids expense of pairwise swappings, and avoids subsampling as in CLARA, CLARANS)
2. Put each point with its closest Medoid (building P-tree component masks as you do this).
3. Repeat 1 & 2 until (some stopping condition – such as: no change in Medoid set?)
4. Can we cluster at step 4 without a scan (create component P-trees??)
A Vertical k-Means Clustering Algorithm
As mentioned on previous slide, a Vertical k-Means algorithm goes similarily:
1. Select k component P-trees (The Goal here is to efficiently get one Ptree mask for each component, e.g., calculate the smallest j : the j-lo gridding has > k cells and agglomerate into precisely k components (by ORing Ptrees of cells with closest means…)
2. Calculate the Means of each component, Ch, by
1. ANDing each basic Ptree with PCi
2. In dimension, Ak , calculate the k-component of the mean as i=bk..0 2i *
rc(PCh ^ Pk,i )
3. Put each pt with closest Mean (building P-tree component masks as you do this).
4. Repeat 2, 3 until (some stopping condition – such as: no change in Mean set?)
5. Can we cluster at step 4 without a scan (create component P-trees??)
Zillions of vertical hybrid clustering algs leap to mind (involving partitioning, hierarchical, density methods)! Pick one!
Finding Density Attractors for Density Clustering Alg
Finding density attractors (and their attractor sets – which, when agglomerated via a density threshold constitutes the generic density clustering algorithm)
1. Pick a point.2. Build out (HOBBit or Ordinary or?) rings one at a time until
the first k neighbors are found.3. In that ring, computer the medoid points (as in step 1 of the
V-k-Medoids algorithm above)4. if the Medoid increases the density, climb to it and goto 25. Else declare it a density attractor and goto 1
j-grids - P-tree relationship1-hi-gridding of R(A1, A2, A3).Bitwidths: 3,2,3 (dim cardinalities: 8,4,8)Using tree node-like identifierscell (tree) id (ci) of the form, c0c1…cd
point (coord) id (pi) of the form, p1,p2,p3
00.0.0000.0.01
00.1.0000.1.01
01.0.0001.0.01
01.1.0101.1.01
00.0.10
00.0.11
00.1.1000.1.11
01.0.10
01.0.11
01.1.10
01.1.11 10.0.0010.0.01
10.1.0010.1.01
10.0.10 10.0.11
10.1.1010.1.11 11.0.00
11.0.01
11.1.0011.1.01
11.0.1011.0.11
11.1.1011.1.11A1
A2
A3
ci=000
ci=001
ci=010
ci=011
ci=100
ci=101
ci=110
ci=111
root
000 001 010 011 101100 110 111
00.0.00
00.0.01
00.1.00
00.1.0101.0.00
01.0.01
01.1.00
01.1.01
00.0.10
00.0.11
00.1.10
00.1.11
01.0.10
01.0.11
01.1.10
01.1.11
10.0.00
10.0.01
10.1.00
10.1.01
11.0.00
11.0.01
11.1.00
11.1.01
10.0.10
10.0.11
10.1.10
10.1.11
11.0.10
11.0.11
11.1.10
11.1.11
A 1-hi-grid yields a P-tree with level-0 (cell level) fanout of 23 and level-1 (point level) fanout of 25. If leaves are segment labelled (not coords):
000.00
000.01
010.00
010.01000.10
000.11
010.10
010.11
. . .
j-grids - P-tree relationship (Cont.)One can view a standard P-tree as nested 1-hi-griddings, with compressed out constant subtrees.R(A1, A2, A3) with bitwidths: 3,2,3
0001
1011
A1
A2
A3
000
001
010
011
100
101
110
111
root
000 001 010 011 101100 110 111
000
000 001 010 011 101100 110 111
001
010
011100
101
110
111
00 01 11 10
Gridding categorical data?
The following bioinformatics (yeast genome) data used was extracted mostly from the MIPS database (Munich Information center for Protein Sequences)
Left column shows features– treat these with hi-order bit (1 iff gene participates)– There may be more levels of hierarchy (e.g., function:
some genes actually cause the function when they express in sufficient quantities, while others are transcription factors for those primary genes. Primary genes have hi bit on, tf’s have second bit on…)
Right column shows # distinct feature values– Bitmap these
Feature Total Valuespathway 80
EC 622complexes 316function 259
localization 43protein class 191
phenotype 181interactions 6347
Data Representation
gene-by-feature table.
For a categorical feature, we consider each category as a separate attribute or column by bit-mapping it.
The resulting table has a total of – 8039 distinct feature bit vectors (corresponding to “items” in
MBR) for– 6374 yeast genes (corresponding to transactions in MBR)
STING: A Statistical Information Grid Approach
Wang, Yang and Muntz (VLDB’97) The spatial area is divided into rectangular cells There are several levels of cells corresponding to different
levels of resolution
STING: A Statistical Information Grid Approach (2)
– Each cell at a high level is partitioned into a number of smaller cells in the next lower level
– Statistical info of each cell is calculated and stored beforehand and is used to answer queries
– Parameters of higher level cells can be easily calculated from parameters of lower level cell
• count, mean, s, min, max • type of distribution—normal, uniform, etc.
– Use a top-down approach to answer spatial data queries– Start from a pre-selected layer—typically with a small number of
cells– For each cell in the current level compute the confidence interval
STING: A Statistical Information Grid Approach (3)
– Remove the irrelevant cells from further consideration– When finish examining the current layer, proceed to the next lower
level – Repeat this process until the bottom layer is reached– Advantages:
• Query-independent, easy to parallelize, incremental update• O(K), where K is the number of grid cells at the lowest level
– Disadvantages:• All the cluster boundaries are either horizontal or vertical, and
no diagonal boundary is detected
WaveCluster (1998)
Sheikholeslami, Chatterjee, and Zhang (VLDB’98) A multi-resolution clustering approach which applies
wavelet transform to the feature space– A wavelet transform is a signal processing technique that
decomposes a signal into different frequency sub-band. Both grid-based and density-based Input parameters:
– # of grid cells for each dimension– the wavelet, and the # of applications of wavelet transform.
WaveCluster (1998)
How to apply wavelet transform to find clusters– Summaries the data by imposing a multidimensional
grid structure onto data space– These multidimensional spatial data objects are
represented in a n-dimensional feature space– Apply wavelet transform on feature space to find the
dense regions in the feature space– Apply wavelet transform multiple times which result in
clusters at different scales from fine to coarse
What Is Wavelet (2)?
Quantization
Transformation
WaveCluster (1998) Why is wavelet transformation useful for clustering
– Unsupervised clustering It uses hat-shape filters to emphasize region where points cluster, but
simultaneously to suppress weaker information in their boundary – Effective removal of outliers– Multi-resolution– Cost efficiency
Major features:– Complexity O(N)– Detect arbitrary shaped clusters at different scales– Not sensitive to noise, not sensitive to input order– Only applicable to low dimensional data
CLIQUE (Clustering In QUEst)
Agrawal, Gehrke, Gunopulos, Raghavan (SIGMOD’98).
http://portal.acm.org/citation.cfm?id=276314
Automatically identifying subspaces of a high dimensional data space that allow better clustering than original space
CLIQUE can be considered as both density-based and grid-based– It partitions each dimension into the same number of equal length interval– It partitions an m-dimensional data space into non-overlapping rectangular
units– A unit is dense if the fraction of total data points contained in the unit
exceeds the input model parameter– A cluster is a maximal set of connected dense units within a subspace
CLIQUE: The Major Steps
Partition the data space and find the number of points that lie inside each cell of the partition.
Identify the subspaces that contain clusters using the Apriori principle
Identify clusters:– Determine dense units in all subspaces of interests– Determine connected dense units in all subspaces of interests.
Generate minimal description for the clusters– Determine maximal regions that cover a cluster of connected
dense units for each cluster– Determination of minimal cover for each cluster
ABSTRACTData mining applications place special requirements on clustering algorithms including:
the ability to find clusters embedded in subspaces of high dimensional data,
scalability, end-user comprehensibility of the results,
non-presumption of any canonical data distribution, and
insensitivity to the order of input records.
We present CLIQUE, a clustering algorithm that satisfies each of these requirements. CLIQUE identifies dense clusters in subspaces of maximum dimensionality.
It generates cluster descriptions in the form of DNF expressions that are minimized for ease of comprehension.
It produces identical results irrespective of the order in which input records are presented and does not presume any specific mathematical form for data distribution.
Through experiments, we show that CLIQUE efficiently finds accurate cluster in large high dimensional datasets.
CLIQUE
Sala
ry
(10,
000)
20 30 40 50 60age
54
31
26
70
20 30 40 50 60age
54
31
26
70
Vac
atio
n(w
eek)
age
Vac
atio
n
Salary 30 50
= 3
Strength and Weakness of CLIQUE
Strength – It automatically finds subspaces of the highest dimensionality such
that high density clusters exist in those subspaces– It is insensitive to the order of records in input and does not
presume some canonical data distribution– It scales linearly with the size of input and has good scalability as
the number of dimensions in the data increases Weakness
– The accuracy of the clustering result may be degraded at the expense of simplicity of the method
Model-Based Clustering Methods
Attempt to optimize the fit between the data and some mathematical model
Statistical and AI approach– Conceptual clustering
• A form of clustering in machine learning• Produces a classification scheme for a set of unlabeled objects• Finds characteristic description for each concept (class)
– COBWEB (Fisher’87) • A popular a simple method of incremental conceptual learning• Creates a hierarchical clustering in the form of a classification tree• Each node refers to a concept and contains a probabilistic description of that
concept
COBWEB Clustering Method
A classification tree
More on Statistical-Based Clustering
Limitations of COBWEB– The assumption that the attributes are independent of each other is
often too strong because correlation may exist– Not suitable for clustering large database data – skewed tree and
expensive probability distributions CLASSIT
– an extension of COBWEB for incremental clustering of continuous data
– suffers similar problems as COBWEB AutoClass (Cheeseman and Stutz, 1996)
– Uses Bayesian statistical analysis to estimate the number of clusters
– Popular in industry
Other Model-Based Clustering Methods
Neural network approaches– Represent each cluster as an exemplar, acting as a
“prototype” of the cluster– New objects are distributed to the cluster whose exemplar
is the most similar according to some dostance measure Competitive learning
– Involves a hierarchical architecture of several units (neurons)
– Neurons compete in a “winner-takes-all” fashion for the object currently being presented
Model-Based Clustering Methods
Self-organizing feature maps (SOMs)
Clustering is also performed by having several units competing for the current object
The unit whose weight vector is closest to the current object wins
The winner and its neighbors learn by having their weights adjusted
SOMs are believed to resemble processing that can occur in the brain
Useful for visualizing high-dimensional data in 2- or 3-D space
Hybrid Clustering
One of the common approaches is to combine k-means method and hierarchical clustering
First partition the dataset into k small clusters, and then merge the clusters based on similarity using hierarchical method.
Hybrid clustering combines partitioning clustering and hierarchical clustering approach
K = 7
Problems of Existing Hybrid Clustering
Predefine the number of preliminary clusters, K.Unable to handle noisy data
0
50
100
150
200
250
300
350
400
450
500
0 100 200 300 400 500 600 7000
20
40
60
80
100
120
0 20 40 60 80 100 120
Similarity Measures
Similarity is fundamental to the definition of a cluster.
1. Minkowski Metric 2. Context Metrics:
1. Mutual neighbor distance2. Conceptual similarity measure
Minkowski Metric Minkowski Metric = ||xi – xj||p It works well when a data set has “compact” or
“isolated” clusters The drawback is the tendency of the largest-scaled
feature to dominate the others. Solutions to this problem include normalization of the
continuous features (to a common range or variance) or other weighting schemes.
Mutual neighbor distance (MND) MND comes from real life observations and is based on a non-
symmetric similarity (e.g., friendship level). Two persons A and B group together as close friends if they
mutually feel that the other is his closest friend. If A feels that B is not such a close friend to him, then even
though B may feel that A is his closest friend, the bond of friendship between them is comparatively weak.
The strength of the bond of friendship between two persons is a function of mutual feeling rather than one-way feeling.
Mutual neighbor distance (MND)
MND(xi, xj) = NN(xi, xj) + NN(xj, xi)
where NN(xi, xj) is the neighbor-ness value (or similarity) of xj to xi
The MND is not a metric (it does not satisfy the triangle inequality). It satisfies the first two conditions of a metric: 1. MND(xi, xj) >= 0, and MND(xi, xj) = 0 xi = xj positive definite2. MND(xi, xj) = MND(xj, xi) symmetric
– Note the symmetry was manufactured by defining MND as the sum of the two non-symmetric NN measures.
In spite of this, MND has been successfully applied in several clustering applications.
This observation supports the viewpoint that the dissimilarity does not need to be a metric.
MND Distance (Cont.)
Figure 4: – NN(A, B) = 1, NN(B, A) = 1, MND(A,B) = 2– NN(B, C) = 1, NN(C, B) = 2, MND(B, C) = 3
Figure 5:– NN(A, B) = 1, NN(B, A) = 4, MND(A,B) = 5– NN(B, C) = 1, NN(C, B) = 2, MND(B, C) = 3
Conceptual Similarity Measure In the case of conceptual clustering, the similarity
between xi and xj is defined as
where is the context (the set of surrounding points, and is a set of pre-defined concepts.
The conceptual similarity measure is the most general similarity measure.
s( xi, xj ) = f( xi, xj, σ, ε )
Conceptual Similarity Measure (Cont.)
The Euclidean distance between points A and B is less than that between B and C.
However, B and C can be viewed as “more similar” than A and B because B and C belong to the same concept (ellipse) and A belongs to a different concept (rectangle).
Problems and ChallengesSome progress has been made in scalable clustering methods
– Partitioning: k-means, k-medoids, CLARANS– Hierarchical: BIRCH, CURE– Density-based: DBSCAN, CLIQUE, OPTICS– Grid-based: STING, WaveCluster– Model-based: Autoclass, Denclue, Cobweb
Current clustering techniques do not address all the requirements adequately Constraint-based clustering analysis: Constraints exist in data space
(bridges and highways) or in user queries Still the curse of cardinality stands as
THE difficult problem!
What Is Outlier Discovery? What are outliers?
– The set of objects are considerably dissimilar from the remainder of the data
– Example: Sports: Michael Jordon, Wayne Gretzky, ... Problem
– Find top n outlier points Applications:
– Credit card fraud detection– Telecom fraud detection– Customer segmentation– Medical analysis
Outlier Discovery: Statistical Approaches
Assume a model underlying distribution that generates data set (e.g. normal distribution)
Use discordancy tests depending on – data distribution– distribution parameter (e.g., mean, variance)– number of expected outliers
Drawbacks– most tests are for single attribute– In many cases, data distribution may not be known
Outlier Discovery: Distance-Based Approach
Introduced to counter the main limitations imposed by statistical methods– We need multi-dimensional analysis without knowing data distribution.
Distance-based outlier: A DB(p, D)-outlier is an object O in a dataset T such that at least fraction p (usually ~99/100 or so) of the objects in T lie at a distance greater than D from O (basically: Too many of the points lie far from O)
A dual formulation is, basically: Too few of the points lie close to O, so O is a DB(p, D)-outlier if at least fraction p (usually ~1/100 or so) of the objects in T lie at a distance less than D from O
»
Algorithms for mining distance-based outliers – Index-based algorithm– Nested-loop algorithm – Cell-based algorithm
Outlier Discovery: Deviation-Based Approach
Identifies outliers by examining the main characteristics of objects in a group
Objects that “deviate” from this description are considered outliers
sequential exception technique – simulates the way in which humans can distinguish unusual objects from
among a series of supposedly like objects
OLAP data cube technique– uses data cubes to identify regions of anomalies in large
multidimensional data
Outlier Detection
Mining rare events, deviant objects and exceptions Find anomaly interests in many domains:
• Criminal activities in electronic commerce• Intrusion in network• Pest infestation in agriculture• Detecting bugs in software, etc
An example in business• British Telecom broke a multimillion dollar fraud
Huge amount of Data Unexpected Knowledge/Anomaly Interest
Outlier Detection
Outlier Analysis (Cont.)
Papadimitriou’s LOCI
(ICDE 2002)Breunig’s LOF
(SIGMOD 2000)Depth
(Preparata 1988)Distribution
(Barnett 1994)
Distance-based Density-basedStatistic-based
Outlier Analysis
Angiulli(TKDE2005)
Ramaswamy(MOD 2000)
Knorr and Ng(VLDB 1998)
Cluster-based
Jiang’s MST(PRL)
He’s CBLOF(PRL)
Distance-based Outlier Detection
Distance-based outlier definition, DB (p, D), by Knorr and Ng (1997)– p: a number percentage; D: distance threshold– Unify distribution-based outlier definitions– Most well known definition
Outlier detection methods– Nested loop method (Knorr and Ng, 1998)– Cell structure-based method (Knorr and Ng, 1998)– Partition-based method (Ramaswamy, 2000)– Work well for small datasets– Not efficient for large datasets
Our proposed work followed aims to improve the efficiency of the DB outlier detection process
Knorr and Ng’s Definition– x is an outlier if {y X |d (y-x) ≥ D & |y| ≥ p*|X|}– d(y,x): distance between y and x
Outlier Definition
Equivalent Description– x is an outlier if {y X |d (y-x) < D & |y|<p*|X|}
·Dx
|y| ≥ p*|X|
·Dx
|y| < (1-p)*|X|
Figure 1. Knorr’s definition Figure 2. Equivalent definition
complement
Definition 1: Disk neighborhood– Define the Disk-neighborhood of x with radius r as a set
satisfying DiskNbr (x, r) = {y X |d(y-x) r}
where, y is any point in DiskNbr(x,r) and d is the distance between y and x.
– Define points in DiskNbr (x, r) as r-neighbors of x.
Definition 2: DiskNbr-based outlier– A point x is considered as a DB (p, D) outlier iff |
DiskNbr(x,D)|≤ (1-p)*|X|.where |DiskNbr (x,D)| is the number of D-neighbors of x,|X| is the total size of dataset X
– (1-p)*|X| is a number threshold given user defined p.
Our Definition
Proposition 1
If |DiskNbr (x, D/2)| ≥ (1-p)*|X|, then all the D/2-neighbors of x are not outliers. Proof: For q DiskNbr(x,D/2)
– DiskNbr (q, D) DiskNbr (x, D/2)– |DiskNbr(q, D)| ≥ |DiskNbr (x, D/2)| ≥ (1-p)*|X|
q is not an outlier. Mark D/2-neighbors as non-outliers A pruning rule
D/2
D
DiskNbr (x, D/2)
qx
D
Figure 3. D/2-neighbors of x marked as non-outliers
Proposition 2
If |DiskNbr (x, 2D)| < (1-p)* |X|, then all D-neighbors of x are outliers.Proof: For q DiskNbr (x, D)
– DiskNbr (q, D) DiskNbr (x, 2D) – |DiskNbr (q, D)| ≤ |DiskNbr(x, 2D)| < (1-p)* |X|
q is an outlier All points in DiskNbr(x,D) are outliers Identify outliers by-neighborhood, faster than by point
D
2D
DiskNbr(x,D)
xq
D
Figure 4. All D-Neighbors of x are outliers
The DBODLP Algorithm
Step 1: Select a point x arbitrarily from X; Step 2: Search D-neighbors of x and calculate |
DiskNbr(x,D)|:
– |DiskNbr(x,D)| < (1-p)*|X| • r <- 2D• Compute |DiskNbr (x, 2*D)| • IF |DiskNbr(x,2*D)| < (1-p)*|X| insert D-neighbors into the outlier set.
ELSE only x is an outlier.
– |DiskNbr(x,D)| ≥ (1-p)*|X|• r <- D/2; • Calculate |DiskNbr(x,D/2)|.• IF |DiskNbr(x,D/2)| ≥ (1-p) * |X|
mark all D/2-neighbors as non-outliers.ELSE only x is non outlier.
Step 3: The process is conducted iteratively until all points in dataset X are examined.
Figure 5. An example
·D
·D
·D/2
x
x
·2D
The DBODLP Algorithm
The algorithm is implemented based on a vertical storage and index model
Vertical Data Structures
Vertical DBODLP
DBODLP is implemented based on P-Tree structure
Predicate Range Tree is used to calculate the number of disk nearest neighbors – PDiskNbr(x,D)= Py>x-D AND Py≤x+D
– |DiskNbr(x,D)|=COUNT (PDiskNbr(x,D))
The experiments are done in DataMIME
Architecture for the DataMIME™ System
(DataMIMEtm = P-tree based Data base and data mining system)
DII (Data Integration Interface)
Data Integration Language DIL
YOUR DATA
Data Repositorylossless, compressed, distributed, vertically-
structured P-tree database
DMI (Data Mining Interface)
YOUR DATA MINING
Ptree (Predicates) Query Language
Query and Data mining
P-Tree Feeders
Meta Data File
Preliminary experimental analysis
0
500
1000
1500
2000
Run Time (in second)
Data Size
Comparison of NL,PODM, PODMP
NL 0.06 0.8 20.03 336.5 1657.9
PODM 0.12 1.15 7.32 65.4 803.2
PODMP 0.11 1.05 5.3 30.2 267.52
256 1024 4096 16384 65536
n
Figure 6. Run Time Comparison
1. Our method has almost an order of magnitude of improvement over the nested loop method in terms of run time
2. All methods produce same outlier set
3. Our method scales better than others
scalability of NL, PODM and PODMP
-200
0
200
400
600
800
1000
1200
1400
1600
1800
256 1024 4096 16384 65536
data size
run
time NL
PODM
PODMP
Figure 7. Scalability Comparisons
Summary Cluster analysis groups objects based on their similarity and
has wide applications Measure of similarity can be computed for various types of
data Clustering algorithms can be categorized into partitioning
methods, hierarchical methods, density-based methods, grid-based methods, and model-based methods
Outlier detection and analysis are very useful for fraud detection, etc. and can be performed by statistical, distance-based or deviation-based approaches
There are still lots of research issues on cluster analysis, such as constraint-based clustering
References (1)
R. Agrawal, J. Gehrke, D. Gunopulos, and P. Raghavan. Automatic subspace clustering of high dimensional data for data mining applications. SIGMOD'98
M. R. Anderberg. Cluster Analysis for Applications. Academic Press, 1973. M. Ankerst, M. Breunig, H.-P. Kriegel, and J. Sander. Optics: Ordering points to identify the
clustering structure, SIGMOD’99. P. Arabie, L. J. Hubert, and G. De Soete. Clustering and Classification. World Scietific, 1996 M. Ester, H.-P. Kriegel, J. Sander, and X. Xu. A density-based algorithm for discovering clusters in
large spatial databases. KDD'96. M. Ester, H.-P. Kriegel, and X. Xu. Knowledge discovery in large spatial databases: Focusing
techniques for efficient class identification. SSD'95. D. Fisher. Knowledge acquisition via incremental conceptual clustering. Machine Learning, 2:139-
172, 1987. D. Gibson, J. Kleinberg, and P. Raghavan. Clustering categorical data: An approach based on
dynamic systems. In Proc. VLDB’98. S. Guha, R. Rastogi, and K. Shim. Cure: An efficient clustering algorithm for large databases.
SIGMOD'98. A. K. Jain and R. C. Dubes. Algorithms for Clustering Data. Printice Hall, 1988.
References (2) L. Kaufman and P. J. Rousseeuw. Finding Groups in Data: an Introduction to Cluster Analysis.
John Wiley & Sons, 1990. E. Knorr and R. Ng. Algorithms for mining distance-based outliers in large datasets. VLDB’98. G. J. McLachlan and K.E. Bkasford. Mixture Models: Inference and Applications to Clustering.
John Wiley and Sons, 1988. P. Michaud. Clustering techniques. Future Generation Computer systems, 13, 1997. R. Ng and J. Han. Efficient and effective clustering method for spatial data mining. VLDB'94. E. Schikuta. Grid clustering: An efficient hierarchical clustering method for very large data sets.
Proc. 1996 Int. Conf. on Pattern Recognition, 101-105. G. Sheikholeslami, S. Chatterjee, and A. Zhang. WaveCluster: A multi-resolution clustering
approach for very large spatial databases. VLDB’98. W. Wang, Yang, R. Muntz, STING: A Statistical Information grid Approach to Spatial Data
Mining, VLDB’97. T. Zhang, R. Ramakrishnan, and M. Livny. BIRCH : an efficient data clustering method for very
large databases. SIGMOD'96.