Transcript Talk Slides - Department of Computer and Information Science and

Clustering and Partitioning for
Spatial and Temporal Data Mining
Vasilis Megalooikonomou
Data Engineering Laboratory (DEnLab)
Dept. of Computer and Information Sciences
Temple University
Philadelphia, PA
www.cis.temple.edu/~vasilis
V. Megalooikonomou, Temple University
Outline
• Introduction
– Motivation – Problems:
• Spatial domain
• Time domain
– Challenges
• Spatial data
– Partitioning and Clustering
– Detection of discriminative patterns
– Results
• Temporal data
– Partitioning
– Vector Quantization
– Results
• Conclusions - Discussion
V. Megalooikonomou, Temple University
Introduction
• Large spatial and temporal databases
• Meta-analysis of data pooled from multiple
studies
• Goal: To understand patterns and discover
associations, regularities and anomalies in
spatial and temporal data
V. Megalooikonomou, Temple University
Problem
Spatial Data Mining:
Given a large collection of spatial data, e.g., 2D or 3D
images, and other data, find interesting things, i.e.:
• associations among image data or among image
and non-image data
• discriminative areas among groups of images
• rules/patterns
• similar images to a query image (queries by
content)
V. Megalooikonomou, Temple University
Challenges
•
•
•
•
•
•
•
•
•
How to apply data mining techniques to images?
Learning from images directly
Heterogeneity and variability of image data
Preprocessing (segmentation, spatial normalization, etc)
Exploration of high correlation between neighboring
objects
Large dimensionality
Complexity of associations
Efficient management of topological/distance information
Spatial knowledge representation / Spatial Access
Methods (SAMs)
V. Megalooikonomou, Temple University
Example: Association Mining – Spatial Data
• Discover associations among spatial and non-spatial data:
• Images {i1, i2,…, iL}
• Spatial regions {s1, s2,…, sK}
• Non-spatial variables {c1, c2,…, cM}
i1
i2
c2
c1 c7
V. Megalooikonomou, Temple University
i3
c2
i4
i5
c1 c3 c9
c6
i6
i7
Example: fMRI contrast maps
Control
V. Megalooikonomou, Temple University
Patient
Applications
Medical Imaging, Bioinformatics, Geography, Meteorology, etc..
V. Megalooikonomou, Temple University
Voxel-based Analysis
• No model on the image data
• Each voxel’s changes analyzed independently - a map of
statistical significance is built
• Discriminatory significance measured by statistical tests (ttest, ranksum test, F-test, etc)
• Statistical Parametric Mapping (SPM)
• Significance of associations measured by chi-squared test,
Fisher’s exact test (a contingency table for each pair of vars)
• Cluster voxels by findings
V. Megalooikonomou, Temple University
[V. Megalooikonomou, C. Davatzikos, E. Herskovits, SIGKDD 1999]
Analysis by grouping of voxels
• Grouping of voxels (atlas-based)
• Prior knowledge increases sensitivity
• Data reduction: 107 voxels R regions (structures)
• Map a ROI onto at least one region
• As good as the atlas being used
• M non-spatial variables, R regions
• Analysis
• Categorical structural variables
• M x R contingency tables, Chi-square/Fisher exact test
• multiple comparison problem
• log-linear analysis, multivariate Bayesian
• Continuous structural variables
• Logistic regression, Mann-Whitney
V. Megalooikonomou, Temple University
Dynamic Recursive Partitioning
• Adaptive partitioning of a 3D volume
V. Megalooikonomou, Temple University
Dynamic Recursive Partitioning
• Adaptive partitioning of a 3D volume
• Partitioning criterion:
discriminative power of feature(s) of hyper-rectangle and
size of hyper-rectangle
V. Megalooikonomou, Temple University
Dynamic Recursive Partitioning
• Adaptive partitioning of a 3D volume
• Partitioning criterion:
discriminative power of feature(s) of hyper-rectangle and
size of hyper-rectangle
V. Megalooikonomou, Temple University
Dynamic Recursive Partitioning
• Adaptive partitioning of a 3D volume
• Partitioning criterion:
discriminative power of feature(s) of hyper-rectangle and
size of hyper-rectangle
V. Megalooikonomou, Temple University
Dynamic Recursive Partitioning
• Adaptive partitioning of a 3D volume
• Partitioning criterion:
discriminative power of feature(s) of hyper-rectangle and
size of hyper-rectangle
• Extract features from discriminative regions
• Reduce multiple comparison problem
(# tests = # partitions < # voxels)
• tests downward closed
V. Megalooikonomou, Temple University
[V. Megalooikonomou, D. Pokrajac, A. Lazarevic, and Z. Obradovic,
SPIE Conference on Visualization and Data Analysis, 2002]
Other Methods for Spatial Data Classification
Distinguishing among distributions:
•Distributional Distances:
- Mahalanobis distance
- Kullback-Leibler divergence (parametric, nonparametric)
•Maximum Likelihood:
- Estimate probability densities and compute likelihood
•EM (Expectation-Maximization) method to model
spatial regions using some base function (Gaussian)
•Static partitioning:
*
• Reduction of the # of attributes as compared to voxel-wise
*
analysis
*
*
• Space partitioned into 3D hyper-rectangles (variables: properties * *
of voxels inside hyper-rectangles) - incrementally increase
*
*
*
discretization
D. Pokrajac, V. Megalooikonomou, A. Lazarevic, D. Kontos, Z. Obradovic,
V. Megalooikonomou, Temple University
Artificial Intelligence in Medicine, Vol. 33, No. 3, pp. 261-280, Mar. 2005.
Experimental Results
Method
Classification Accuracy (%)
Criterion
Threshold
Tree depth
Controls
Patients
Total
correlation
0.4
3
82
93
88
0.05
3
89
100
94
0.05
4
84
100
92
0.01
4
87
100
93
0.05
3
87
100
93
0.05
4
80
100
90
0.01
4
87
96
91
Maximum Likelihood / EM
77
67
72
Maximum Likelihood / k-means
77
83
80
Kullback-Leibler / EM
79
57
68
Kullback-Leibler / k-means
77
66
71
t-test
DRP
ranksum
V. Megalooikonomou, Temple University
Areas discovered by DRP with t-test: significance
threshold=0.05, maximum tree depth=3. Colorbar shows
significance
Comparison of number of tests performed
Number of tests
Thresh
.
0.05
Depth
DRP
Voxel Wise
3
569
201774
0.05
4
4425
201774
0.01
4
4665
201774
[D. Kontos, V. Megalooikonomou, D. Pokrajac, A. Lazarevic, Z. Obradovic,
O. B. Boyko, J. Ford, F. Makedon, A. J. Saykin, MICCAI 2004]
Experimental Results
Discriminative sub-regions
detected when applying (a)
DRP and (b) voxel-wise
analysis with ranksum test
and significance threshold
0.05 to the real fMRI
volume data
(a)
(b)
Impact:
• Assist in interpretation of images (e.g., facilitating diagnosis)
• Enable researchers to integrate, manipulate and analyze large
volumes of image data
V. Megalooikonomou, Temple University
Time Sequence Analysis
Time Sequence: A sequence
(ordered collection) of real
values: X = x1, x2,…, xn
• Time series data abound in many applications …
• Challenges:
– High dimensionality
– Large number of sequences
– Similarity metric definition
• Similarity analysis (e.g., find stocks similar to that of IBM)
• Goals: high accuracy, (high speed) in similarity searches among time
series and in discovering interesting patterns
• Applications: clustering, classification, similarity searches,
summarization
V. Megalooikonomou, Temple University
Dimensionality Reduction Techniques
• DFT: Discrete Fourier Transform
• DWT: Discrete Wavelet Transform
• SVD: Singular Value Decomposition
•APCA: Adaptive Piecewise Constant Approximation
• PAA: Piecewise Aggregate Approximation
• SAX: Symbolic Aggregate approXimation
•
•… •
V. Megalooikonomou, Temple University
Similarity distances for time series
• Euclidean Distance:
most common, sensitive to shifts
• Dynamic Time Warping:
improving accuracy but slow: O(n2)
• Envelope-based DTW:
faster: O(n)
A more intuitive idea:
two series should be considered similar if they have enough
non-overlapping time-ordered pairs of subsequences that are
similar (Agrawal et al. VLDB, 1995)
V. Megalooikonomou, Temple University
Partitioning – Piecewise Constant
Approximations
Original time series
(n points)
Piecewise constant approximation (PCA)
or Piecewise Aggregate Approximation
(PAA), [Yi and Faloutsos ’00, Keogh et
al, ’00] (n' segments)
Adaptive Piecewise Constant
Approximation (APCA), [Keogh et al.,
’01] (n" segments)
V. Megalooikonomou, Temple University
Multiresolution Vector Quantized
approximation (MVQ)
Partitions a sequence into equal-length segments and uses VQ to
represent each sequence by appearance frequencies of keysubsequences
1) Uses a ‘vocabulary’ of subsequences (codebook) –
training is involved
2) Takes multiple resolutions into account – keeps both
local and global information
3) Unlike wavelets partially ignores the ordering of
‘codewords’
3) Can exploit prior knowledge about the data
4) Employs a new distance metric
V. Megalooikonomou, Temple University
[V. Megalooikonomou, Q. Wang, G. Li, C. Faloutsos, ICDE 2005]
Methodology
Codebook
l
s=16
s
Generation
Series
cmdbca
minj j a
hldf ko
ogcbl p
l hnkkk
kkgj hh
Transformation
ifaj bb
ma I n j m
phcako
occbl h
pl cacg
gkgj lp
……
V. Megalooikonomou, Temple University
Series
Encoding
1121000000001000
1200010011000000
1000000012001100
1000000011002100
0001010100110010
1010000100100011
……
Methodology
Creating a ‘vocabulary’
Q: How to create?
A: Use Vector Quantization, in
particular, the Generalized Lloyd
Algorithm (GLA)
Frequently appearing
patterns in
subsequences
Output:
A codebook with s codewords
Representing time series
X = x1, x2,…, xn
is encoded with a new representation
f = (f1,f2,…, fs)
(fi is the frequency of the i th codeword in X)
V. Megalooikonomou, Temple University
Methodology
New distance metric:
The histogram model is used to calculate
similarity at each resolution level:
1
S HM (q, t ) 
1  dis (q, t )
wit
h
V. Megalooikonomou, Temple University
s
f i ,t  f i , q
i 1
1  f i ,t  f i , q
dis (q, t )  
fi,t
fi,q
1 2...s
Methodology
Time series summarization:
• High level information (frequently appearing patterns) is
more useful
• The new representation can provide this kind of
information
Both codeword
(pattern) 3 & 5
show up 2 times
V. Megalooikonomou, Temple University
Methodology
Problems of frequency based encoding:
• It is hard to define an approximate resolution
(codeword length)
• It may lose global information
V. Megalooikonomou, Temple University
Methodology
Solution: Use multiple resolutions:
• It is hard to define an approximate resolution
(codeword length)
• It may lose global information
V. Megalooikonomou, Temple University
Methodology
Proposed distance metric:
Weighted sum of similarities, at all
resolution levels
c
S HHM (q, d j )   w i * S HMi (q, d j )
i 1
where c is the number of resolution levels
similarity @ level i
•lacking any prior knowledge equal weights to
all resolution levels works well most of the time
V. Megalooikonomou, Temple University
MVQ: Example of Codebooks
• Codebook for the
first level
• Codebook for the
second level (more
codewords since
there are more
details)
V. Megalooikonomou, Temple University
Experiments
Datasets

SYNDATA (control chart data): synthetic
CAMMOUSE: 3 *5 sequences obtained using the
Camera Mouse Program
 RTT: RTT measurements from UCR to CMU with
sending rate of 50 msec for a day

V. Megalooikonomou, Temple University
Experiments
Best Match Searching:
Matching accuracy: % of knn’s (found by different approaches)
that are in same class
Accuracy 
V. Megalooikonomou, Temple University
| knn(q)  std_set(q) |
 100%
k
Experiments
Best Match Searching
SYNDATA
CAMMOUSE
Method
Weight
Vector
Accuracy
Method
Weight Vector
Accuracy
Single
level
VQ
[1 0 0 0 0]
0.55
[1 0 0 0 0]
0.56
[0 1 0 0 0]
0.70
Single
level
VQ
[0 1 0 0 0]
0.60
[0 0 1 0 0]
0.65
[0 0 1 0 0]
0.44
[0 0 0 1 0]
0.48
[0 0 0 1 0]
0.56
[0 0 0 0 1]
0.46
[0 0 0 0 1]
0.60
[1 1 1 1 1]
0.83
MVQ
[1 1 1 1 1]
0.83
0.51
Euclidean
MVQ
Euclidean
V. Megalooikonomou, Temple University
0.58
Experiments
Best Match Searching
MVQ
MVQ
(a)
(b)
Precision-recall for different methods
(a) on SYNDATA dataset (b) on CAMMOUSE dataset
V. Megalooikonomou, Temple University
Experiments
Clustering experiments
Given two clusterings, G=G1, G2, …, GK (the true clusters),
and A = A1, A2, …, Ak (clustering result by a certain
method), the clustering accuracy is evaluated with the
cluster similarity defined as:
Sim(G, A) 
imax j Sim(Gi , A j )
V. Megalooikonomou, Temple University
k
with Sim(Gi, Aj) 
2 | Gi  A j |
| Gi |  | A j |
[Gavrilov, M., Anguelov, D., Indyk, P. and Motwani, R., KDD 2000]
Experiments
Clustering experiments
SYNDATA
Method
RTT
Method
Weight
Vector
Single level
VQ
[1 0 0 0 0]
0.55
[0 1 0 0 0]
0.52
0.63
[0 0 1 0 0]
0.57
[0 0 0 1 0]
0.51
[0 0 0 1 0]
0.80
[0 0 0 0 1]
0.49
[0 0 0 0 1]
0.79
[1 1 1 1 1]
0.82
MVQ
[0 0 0 1 1]
0.81
DFT
0.67
DFT
0.54
SAX
0.65
SAX
0.54
DTW
0.80
DTW
0.62
Euclidean
0.55
Euclidean
0.50
Single level
VQ
MVQ
Weight
Vector
Accuracy
[1 0 0 0 0]
0.69
[0 1 0 0 0]
0.71
[0 0 1 0 0]
V. Megalooikonomou, Temple University
Accuracy
MVQ: Example: Two Time Series
• Given two time series t1 and t2 as follows:
• In the first level, they are encoded with the same codeword (3), so
they are not distinguishable
• In the second level, more details are recorded. These two series have
different encoded form: the first series is encoded with codeword 1
and 4, the second one is encoded with codewords 9 and 12.
V. Megalooikonomou, Temple University
Analysis of images by projection to 1D
(a)
(b)
(c)
(a) linear mapping of a 3D fMRI scan, (b) effect of binning by representing each bin with its Vmean measurement,
(c) the discriminative voxels after applying the t-test with θ=0.05
•
•
•
•
Hilbert Space Filling Curve
Binning
Statistical tests of significance on groups of points
Identification of discriminative areas by back-projection
V. Megalooikonomou, Temple University
[D. Kontos, V. Megalooikonomou, N. Ghubade, and C. Faloutsos.
IEEE Engineering in Medicine and Biology Society (EMBS), 2003]
Applying time series techniques
(a)
(b)
Areas discovered: (a) θ=0.05, (b) θ=0.01. The colorbar shows significance.
Results: 87%-98% classification accuracy (t-test, CATX)
Variation: Concatenate the values of statistically
significant areas  spatial sequences
• Pattern analysis using the similarity between spatial
sequences and time sequences
• SVD, DFT, DWT, PCA (clustering accuracy:
89-100%)
V. Megalooikonomou, Temple University
[Q. Wang, D. Kontos, G. Li and V. Megalooikonomou, ICASSP 2004]
Conclusions
• ‘Find patterns/interesting things’ efficiently and
robustly in spatial and temporal data
• Use of partitioning and clustering
• Analysis at multiple resolutions
• Reduction of the number of tests performed
• Intelligent exploration of the space to find
discriminative areas
• Reduction of dimensionality
• Symbolic representation
• Nice summarization
V. Megalooikonomou, Temple University
Collaborators
Faculty:
• Zoran Obradovic
• Orest Boyko
• James Gee
• Andrew Saykin
• Christos Faloutsos
• Christos Davatzikos
• Edward Herskovits
• Fillia Makedon
• Dragoljub Pokrajac
V. Megalooikonomou, Temple University
Students:
• Despina Kontos
• Qiang Wang
• Guo Li
Others:
• James Ford
• Alexandar Lazarevic
Thank you!
Acknowledgements
This research has been funded by:
– National Science Foundation CAREER award 0237921
– National Science Foundation Grant 0083423
– National Institutes of Health Grant R01 MH68066 funded by NIMH, NINDS, and NIA
V. Megalooikonomou, Temple University