Data Mining Principles and Algorithms Chapter 10 1

Data Mining: Principles and Algorithms — Chapter 10. 1 — — Mining Object, Spatial, and Multimedia Data— ©Jiawei Han Department of Computer Science University of Illinois at Urbana-Champaign www. cs. uiuc. edu/~hanj 10/27/2020 Data Mining: Principles and Algorithms 1

10/27/2020 Data Mining: Principles and Algorithms 2

Mining Object, Spatial and Multi-Media Data n Mining object data sets n Mining spatial databases and data warehouses n Spatial DBMS n Spatial Data Warehousing n Spatial Data Mining n Spatiotemporal Data Mining n Mining multimedia data n Summary 10/27/2020 Data Mining: Principles and Algorithms 3

Mining Complex Data Objects: Generalization of Structured Data n Set-valued attribute n Generalization of each value in the set into its corresponding higher-level concepts n Derivation of the general behavior of the set, such as the number of elements in the set, the types or value ranges in the set, or the weighted average for numerical data n E. g. , hobby = {tennis, hockey, chess, violin, PC_games} generalizes to {sports, music, e_games} n List-valued or a sequence-valued attribute n Same as set-valued attributes except that the order of the elements in the sequence should be observed in the generalization 10/27/2020 Data Mining: Principles and Algorithms 4

Generalizing Spatial and Multimedia Data n n Spatial data: n Generalize detailed geographic points into clustered regions, such as business, residential, industrial, or agricultural areas, according to land usage n Require the merge of a set of geographic areas by spatial operations Image data: n n n Extracted by aggregation and/or approximation Size, color, shape, texture, orientation, and relative positions and structures of the contained objects or regions in the image Music data: n n 10/27/2020 Summarize its melody: based on the approximate patterns that repeatedly occur in the segment Summarized its style: based on its tone, tempo, or the major musical instruments played Data Mining: Principles and Algorithms 5

Generalizing Object Data n n n Object identifier n generalize to the lowest level of class in the class/subclass hierarchies Class composition hierarchies n generalize only those closely related in semantics to the current one Construction and mining of object cubes n Extend the attribute-oriented induction method n Apply a sequence of class-based generalization operators on different attributes n Continue until getting a small number of generalized objects that can be summarized as a concise in high-level terms n Implementation n Examine each attribute, generalize it to simple-valued data n Construct a multidimensional data cube (object cube) n Problem: it is not always desirable to generalize a set of values to single-valued data 10/27/2020 Data Mining: Principles and Algorithms 6

Ex. : Plan Mining by Divide and Conquer n Plan: a sequence of actions n n Plan mining: extraction of important or significant generalized (sequential) patterns from a planbase (a large collection of plans) n n n E. g. , Travel (flight): <traveler, departure, arrival, d-time, airline, price, seat> E. g. , Discover travel patterns in an air flight database, or find significant patterns from the sequences of actions in the repair of automobiles Method n Attribute-oriented induction on sequence data n n Divide & conquer: Mine characteristics for each subsequence n 10/27/2020 A generalized travel plan: <small-big*-small> E. g. , big*: same airline, small-big: nearby region Data Mining: Principles and Algorithms 7

A Travel Database for Plan Mining n Example: Mining a travel planbase Travel plan table Airport info table 10/27/2020 Data Mining: Principles and Algorithms 8

Multidimensional Analysis n A multi-D model for the planbase Strategy n n n 10/27/2020 Generalize the planbase in different directions Look for sequential patterns in the generalized plans Derive high-level plans Data Mining: Principles and Algorithms 9

Multidimensional Generalization Multi-Dimensional generalization of the planbase Merging consecutive, identical actions in plans 10/27/2020 Data Mining: Principles and Algorithms 10

Generalization-Based Sequence Mining n n Generalize planbase in multidimensional way using dimension tables Use # of distinct values (cardinality) at each level to determine the right level of generalization (level“planning”) Use operators merge “+”, option “[]” to further generalize patterns Retain patterns with significant support 10/27/2020 Data Mining: Principles and Algorithms 11

Generalized Sequence Patterns n Airport. Size-sequence survives the min threshold (after applying merge operator): S-L+-S [35%], L+-S [30%], S-L+ [24. 5%], L+ [9%] n After applying option operator: [S]-L+-[S] [98. 5%] n n Most of the time, people fly via large airports to get to final destination Other plans: 1. 5% of chances, there are other patterns: S-S, L-S-L 10/27/2020 Data Mining: Principles and Algorithms 12

Mining Object, Spatial and Multi-Media Data n Mining object data sets n Mining spatial databases and data warehouses n Spatial DBMS n Spatial Data Warehousing n Spatial Data Mining n Spatiotemporal Data Mining n Mining multimedia data n Summary 10/27/2020 Data Mining: Principles and Algorithms 13

What Is a Spatial Database System? n Geometric, geographic or spatial data: space-related data n Example: Geographic space (2 -D abstraction of earth surface), VLSI design, model of human brain, 3 -D space representing the arrangement of chains of protein molecule. n Spatial database system vs. image database systems. n Image database system: handling digital raster image (e. g. , satellite sensing, computer tomography), may also contain techniques for object analysis and extraction from images and some spatial database functionality. n Spatial (geometric, geographic) database system: handling objects in space that have identity and well-defined extents, locations, and relationships. 10/27/2020 Data Mining: Principles and Algorithms 14

GIS (Geographic Information System) n n Analysis and visualization of geographic data Common analysis functions of GIS n Search (thematic search, search by region) n Location analysis (buffer, corridor, overlay) n Terrain analysis (slope/aspect, drainage network) n Flow analysis (connectivity, shortest path) n Distribution (nearest neighbor, proximity, change detection) n Spatial analysis/statistics (pattern, centrality, similarity, topology) n Measurements (distance, perimeter, shape, adjacency, direction) 10/27/2020 Data Mining: Principles and Algorithms 15

Spatial DBMS (SDBMS) n SDBMS is a software system that supports spatial data models, spatial ADTs, and a query language supporting them n supports spatial indexing, spatial operations efficiently, and query optimization n can work with an underlying DBMS Examples n Oracle Spatial Data Catridge n ESRI Spatial Data Engine 10/27/2020 Data Mining: Principles and Algorithms 16

Modeling Spatial Objects n What needs to be represented? n Two important alternative views n Single objects: distinct entities arranged in space each of which has its own geometric description n n modeling cities, forests, rivers Spatially related collection of objects: describe space itself (about every point in space) n modeling land use, partition of a country into districts 10/27/2020 Data Mining: Principles and Algorithms 17

Modeling Single Objects: Point, Line and Region n Point: location only but not extent n Line (or a curve usually represented by a polyline, a sequence of line segment): n moving through space, or connections in space (roads, rivers, cables, etc. ) n Region: n Something having extent in 2 D-space (country, lake, park). It may have a hole or consist of several disjoint pieces. 10/27/2020 Data Mining: Principles and Algorithms 18

Modeling Spatially Related Collection of Objects n Modeling spatially related collection of objects: plane partitions and networks. n A partition: a set of region objects that are required to be disjoint (e. g. , a thematic map). There exist often pairs of objects with a common boundary (adjacency relationship). n A network: a graph embedded into the plane, consisting of a set of point objects, forming its nodes, and a set of line objects describing the geometry of the edges, e. g. , highways. rivers, power supply lines. n Other interested spatially related collection of objects: nested partitions, or a digital terrain (elevation) model. 10/27/2020 Data Mining: Principles and Algorithms 19

Spatial Data Types and Models Field-based model: raster data n framework: partitioning of space n Object-based model: vector model n point, line, polygon, Objects, Attributes n 10/27/2020 Data Mining: Principles and Algorithms 20

Spatial Query Language Spatial query language n Spatial data types, e. g. point, line segment, polygon, … n Spatial operations, e. g. overlap, distance, nearest neighbor, … n Callable from a query language (e. g. SQL 3) of underlying DBMS SELECT S. name FROM Senator S WHERE S. district. Area() > 300 n Standards n SQL 3 (a. k. a. SQL 1999) is a standard for query languages n OGIS is a standard for spatial data types and operators n Both standards enjoy wide support in industry n 10/27/2020 Data Mining: Principles and Algorithms 21

Spatial Data Types by OGIS 10/27/2020 Data Mining: Principles and Algorithms 22

Query Processing Efficient algorithms to answer spatial queries n Common Strategy: filter and refine n Filter: Query Region overlaps with MBRs (minimum bounding rectangles) of B, C, D n Refine: Query Region overlaps with B, C n 10/27/2020 Data Mining: Principles and Algorithms 23

Join Query Processing Determining Intersection Rectangle n Plane Sweep Algorithm n Place sweep filter identifies 5 intersections for refinement step n 10/27/2020 Data Mining: Principles and Algorithms 24

File Organization and Indices SDBMS: Dataset is in the secondary storage, e. g. disk n Space Filling Curves: An ordering on the locations in a multi-dimensional space n Linearize a multi-dimensional space n Helps search efficiently n 10/27/2020 Data Mining: Principles and Algorithms 25

File Organization and Indices n Spatial Indexing n B-tree works on spatial data with space filling curve n R-tree: Heighted balanced extention of B+ tree Objects are represented as MBR n provides better performance n 10/27/2020 Data Mining: Principles and Algorithms 26

Spatial Query Optimization A spatial operation can be processed using different strategies n Computation cost of each strategy depends on many parameters n n Query optimization is the process of n ordering operations in a query and n selecting efficient strategy for each operation n based on the details of a given dataset 10/27/2020 Data Mining: Principles and Algorithms 27

Spatial Data Warehousing n Spatial data warehouse: Integrated, subject-oriented, time-variant, and nonvolatile spatial data repository n Spatial data integration: a big issue n Structure-specific formats (raster- vs. vector-based, OO vs. relational models, different storage and indexing, etc. ) n n Vendor-specific formats (ESRI, Map. Info, Integraph, IDRISI, etc. ) n Geo-specific formats (geographic vs. equal area projection, etc. ) Spatial data cube: multidimensional spatial database n 10/27/2020 Both dimensions and measures may contain spatial components Data Mining: Principles and Algorithms 28

Dimensions and Measures in Spatial Data Warehouse n Dimensions n n n 10/27/2020 n non-spatial n e. g. “ 25 -30 degrees” generalizes to“hot” (both are strings) spatial-to-nonspatial n e. g. Seattle generalizes to description “Pacific Northwest” (as a string) spatial-to-spatial n e. g. Seattle generalizes to Pacific Northwest (as a spatial region) Measures n n numerical (e. g. monthly revenue of a region) n distributive (e. g. count, sum) n algebraic (e. g. average) n holistic (e. g. median, rank) spatial n collection of spatial pointers (e. g. pointers to all regions with temperature of 25 -30 degrees in July) Data Mining: Principles and Algorithms 29

Spatial-to-Spatial Generalization n n Generalize detailed geographic points into clustered regions, such as businesses, residential, industrial, or agricultural areas, according to land usage Dissolve Requires the merging of a set of geographic areas by spatial operations Intersect 10/27/2020 Merge Clip Union Data Mining: Principles and Algorithms 30

Example: British Columbia Weather Pattern Analysis n Input n n Output n n A map that reveals patterns: merged (similar) regions Goals n n A map with about 3, 000 weather probes scattered in B. C. Daily data for temperature, precipitation, wind velocity, etc. Data warehouse using star schema Interactive analysis (drill-down, slice, dice, pivot, roll-up) Fast response time Minimizing storage space used Challenge n 10/27/2020 A merged region may contain hundreds of “primitive” regions (polygons) Data Mining: Principles and Algorithms 31

Star Schema of the BC Weather Warehouse n Spatial data warehouse n Dimensions n region_name n time n temperature n precipitation n Measurements n region_map n area n count Dimension table 10/27/2020 Data Mining: Principles and Algorithms Fact table 32

Dynamic Merging of Spatial Objects è è Materializing (precomputing) all? —too much storage space On-line merge? —slow, expensive Precompute rough approximations? — accuracy trade off A better way: object-based, selective (partial) materialization 10/27/2020 Data Mining: Principles and Algorithms 33

Methods for Computing Spatial Data Cubes n n On-line aggregation: collect and store pointers to spatial objects in a spatial data cube n expensive and slow, need efficient aggregation techniques Precompute and store all the possible combinations n huge space overhead Precompute and store rough approximations in a spatial data cube n accuracy trade-off Selective computation: only materialize those which will be accessed frequently n a reasonable choice 10/27/2020 Data Mining: Principles and Algorithms 34

Spatial Association Analysis n Spatial association rule: A B [s%, c%] n n n A and B are sets of spatial or non-spatial predicates n Topological relations: intersects, overlaps, disjoint, etc. n Spatial orientations: left_of, west_of, under, etc. n Distance information: close_to, within_distance, etc. s% is the support and c% is the confidence of the rule Examples 1) is_a(x, large_town) ^ intersect(x, highway) ® adjacent_to(x, water) [7%, 85%] 2) What kinds of objects are typically located close to golf courses? 10/27/2020 Data Mining: Principles and Algorithms 35

Progressive Refinement Mining of Spatial Association Rules n n Hierarchy of spatial relationship: n g_close_to: near_by, touch, intersect, contain, etc. n First search for rough relationship and then refine it Two-step mining of spatial association: n Step 1: Rough spatial computation (as a filter) n n Step 2: Detailed spatial algorithm (as refinement) n 10/27/2020 Using MBR or R-tree for rough estimation Apply only to those objects which have passed the rough spatial association test (no less than min_support) Data Mining: Principles and Algorithms 36

Mining Spatial Co-location Rules n Co-location rule is similar to association rule but explore more relying spatial auto-correlation n It leads to efficient processing n It can be integrated with progressive refinement to further improve its performance n Spatial co-location mining idea can be applied to clustering, classification, outlier analysis and other potential mining tasks 10/27/2020 Data Mining: Principles and Algorithms 37

Spatial Autocorrelation n Spatial data tends to be highly self-correlated n Example: Neighborhood, Temperature n Items in a traditional data are independent of each other, whereas properties of locations in a map are often “auto-correlated”. n First law of geography: “Everything is related to everything, but nearby things are more related than distant things. ” 10/27/2020 Data Mining: Principles and Algorithms 38

Spatial Autocorrelation (cont’d) 10/27/2020 Data Mining: Principles and Algorithms 39

Spatial Classification n Methods in classification n Association-based multi-dimensional classification Example: classifying house value based on proximity to lakes, highways, mountains, etc. Assuming learning samples are independent of each other n n Decision-tree classification, Naïve-Bayesian classifier + boosting, neural network, logistic regression, etc. Spatial auto-correlation violates this assumption! Popular spatial classification methods n Spatial auto-regression (SAR) n Markov random field (MRF) 10/27/2020 Data Mining: Principles and Algorithms 40

Spatial Auto-Regression n Linear Regression Y=X + n Spatial autoregressive regression (SAR) Y = WY + X + n W: neighborhood matrix. n models strength of spatial dependencies n error vector The estimates of and can be derived using maximum likelihood theory or Bayesian statistics 10/27/2020 Data Mining: Principles and Algorithms 41

Markov Random Field Based Bayesian Classifiers n n Bayesian classifiers MRF n A set of random variables whose interdependency relationship is represented by an undirected graph (i. e. , a symmetric neighborhood matrix) is called a Markov Random Field. n n n 10/27/2020 Li denotes set of labels in the neighborhood of si excluding labels at si Pr(Ci | Li) can be estimated from training data by examine the ratios of the frequencies of class labels to the total number of locations Pr(X|Ci, Li) can be estimated using kernel functions from the observed values in the training dataset Data Mining: Principles and Algorithms 42

SAR v. s. MRF 10/27/2020 Data Mining: Principles and Algorithms 43

Spatial Trend Analysis n Function n Detect changes and trends along a spatial dimension n Study the trend of non-spatial or spatial data changing with space n Application examples n Observe the trend of changes of the climate or vegetation with increasing distance from an ocean n Crime rate or unemployment rate change with regard to city geo-distribution 10/27/2020 Data Mining: Principles and Algorithms 44

Spatial Cluster Analysis n n Mining clusters—k-means, k-medoids, hierarchical, density-based, etc. Analysis of distinct features of the clusters 10/27/2020 Data Mining: Principles and Algorithms 45

Constraints-Based Clustering n Constraints on individual objects n n Clustering parameters as constraints n n K-means, density-based: radius, min-# of points Constraints specified on clusters using SQL aggregates n n Simple selection of relevant objects before clustering Sum of the profits in each cluster > $1 million Constraints imposed by physical obstacles n 10/27/2020 Clustering with obstructed distance Data Mining: Principles and Algorithms 46

Constrained Clustering: Planning ATM Locations C 2 e g d Bri C 3 C 1 River Mountain Spatial data with obstacles 10/27/2020 C 4 Clustering without taking obstacles into consideration Data Mining: Principles and Algorithms 47

Spatial Outlier Detection n n Outlier n Global outliers: Observations which is inconsistent with the rest of the data n Spatial outliers: A local instability of non-spatial attributes Spatial outlier detection n Graphical tests n Variogram clouds n Moran scatterplots n Quantitative tests n Scatterplots n Spatial Statistic Z(S(x)) n Quantitative tests are more accurate than Graphical tests 10/27/2020 Data Mining: Principles and Algorithms 48

Spatial Outlier Detection─Variogram Clouds n Graphical method n n n For each pair of locations, the square-root of the absolute difference between attribute values at the locations versus the Euclidean distance between the locations are plotted Nearby locations with large attribute difference indicate a spatial outlier Quantitative method n Compute spatial statistic Z(S(x)) 10/27/2020 Data Mining: Principles and Algorithms 49

Spatial Outlier Detection—Moran Scatterplots n Graphical tests n A plot of normalized attribute value Z against the neighborhood average of normalized attribute values (W • Z) The upper left and lower right quadrants indicate a spatial outlier Computation method n Fit a linear regression line n Select points (e. g. P, Q, S) which are from the regression line greater than specified residual error n n 10/27/2020 Data Mining: Principles and Algorithms 50

Mining Spatiotemporal Data n n Spatiotemporal data n Data has spatial extensions and changes with time n Ex: Forest fire, moving objects, hurricane & earthquakes Automatic anomaly detection in massive moving objects n Moving objects are ubiquitous: GPS, radar, etc. n Ex: Maritime vessel surveillance n Problem: Automatic anomaly detection 10/27/2020 Data Mining: Principles and Algorithms 51

Analysis: Mining Anomaly in Moving Objects n n n Raw analysis of collected data does not fully convey “anomaly” information More effective analysis relies on higher semantic features Examples: n A speed boat moving quickly in open water n A fishing boat moving slowly into the docks n A yacht circling slowly around landmark during night hours 10/27/2020 Data Mining: Principles and Algorithms 52

Framework: Motif-Based Feature Analysis n n n Motif-based representation n A motif is a prototypical movement pattern n View a movement path as a sequence of motif expressions Motif-oriented feature space n Automated motif feature extraction n Semantic-level features Classification n Anomaly detection via classification n High dimensional classifier 10/27/2020 Data Mining: Principles and Algorithms 53

Movement Motifs n Prototypical movement of object n n Right-turn, U-turn Can be either defined by an expert or discovered automatically from data n Defined in our framework Extracted in movement paths Path becomes a set of motif expressions 10/27/2020 Data Mining: Principles and Algorithms 54

Motif Expression Attributes n n n Each motif expression has attributes (e. g. , speed, location, size) Attributes express how a motif was expressed Conveys semantic information useful for classification n n 10/27/2020 a tight circle at 30 mph near landmark Y. A tight circle at 10 mph in location X Data Mining: Principles and Algorithms 55

Motif-Oriented Feature Space n n n Attributes describe how motifs are expressed Let there be A attributes, each path is a set of (A+1)-tuples {(mi, v 1, v 2, …, v. A), (mj, v 1, v 2, …, v. A)} Naïve Feature space construction n Let each distinct (mj, v 1, v 2, …, v. A) be a feature n If path exhibits a particular motif-expression, its value is 1. Otherwise, its value is 0. 10/27/2020 Data Mining: Principles and Algorithms 56

Analyzing Naïve Feature Space n n Let there be M distinct motifs and V different possible values for each of the A attributes Size of feature space is M * VA n V is usually very large due to high granularity of measurements n n E. g. , seconds for time or meters for location Modest values for A and M could lead to extremely high dimensional feature space 10/27/2020 Data Mining: Principles and Algorithms 57

More on Naïve Feature Space n n High dimensional feature space could make effective learning hard More importantly, high granular features make generalization impossible! n n n Intuition: should have features that describe general high-level concepts n n n (mj, v 1, 10: 01 am, …, v. A) vs (mj, v 1, 10: 02 am, …, v. A) Learning on one feature has no effect on another feature “Early Morning” instead of 2: 03 am, 2: 04 am, … “Near Location X” instead of “ 50 m west of Location X” Solution: Clustering on naïve feature space 10/27/2020 Data Mining: Principles and Algorithms 58

Motif Feature Extraction n For each motif attribute, cluster values to form higher level concepts Frequency and distribution in learning data dictates the final clusters Hierarchical micro-clustering n Small clusters so concepts are not merged unnecessarily n Hierarchy allows flexibility in describing objects n For example: “afternoon” vs. “early afternoon” and “late afternoon” 10/27/2020 Data Mining: Principles and Algorithms 59

Feature Clustering n n n Rough, fast micro-clustering method based on BIRCH (SIGMOD’ 96) A micro-cluster is represented by a CF Vector: CF = (n, LS, SS) Centroid and radius can be calculated from CF vector CF Additive Theorem allows two CF Vectors to be combined quickly and losslessly CF Tree is a hierarchy of CF Vectors n n 10/27/2020 A parent CF Vector holds information for all descendent CF Vectors Leaf CF Vector corresponds to a set of actual points Data Mining: Principles and Algorithms 60

More on Feature Clustering n n n Build CF Tree from raw data, much like B-tree Two parameters in clustering n B: branching factor of CF Tree n T: radius threshold of CF Vector Parameters control how fine micro-clusters are constructed Hierarchical agglomerative clustering on leaves of CF Tree Entire process is efficient: O(N) 10/27/2020 Data Mining: Principles and Algorithms 61

Extracted Feature Space n n Leaf nodes in final clustering become the new features More general than the original naïve feature space n Dimensionality could still be moderately high n Use Support Vector Machine for classification 10/27/2020 Data Mining: Principles and Algorithms 62

Experiments n Synthetic Data n n n Generated at motif-expression level Abnormal paths are injected with abnormal motif-expressions Classifiers n n 10/27/2020 SVM using naïve feature space SVM using extracted feature spaces of varying refinement levels Data Mining: Principles and Algorithms 63

Experiment 10/27/2020 Data Mining: Principles and Algorithms 64

Experiment (2) 10/27/2020 Data Mining: Principles and Algorithms 65

Summary: Moving Object Anomaly Detection n Higher level semantic analysis of moving objects yields better results Automated feature extraction Future work n Automatic determination of t parameter n Better use of feature space hierarchy n Other analysis, such as clustering and local outlier detection for anomaly detection n Mining other knowledge for moving objects 10/27/2020 Data Mining: Principles and Algorithms 66

Mining Object, Spatial and Multi-Media Data n Mining object data sets n Mining spatial databases and data warehouses n Spatial DBMS n Spatial Data Warehousing n Spatial Data Mining n Spatiotemporal Data Mining n Mining multimedia data n Summary 10/27/2020 Data Mining: Principles and Algorithms 67

Similarity Search in Multimedia Data n Description-based retrieval systems n n Build indices and perform object retrieval based on image descriptions, such as keywords, captions, size, and time of creation n Labor-intensive if performed manually n Results are typically of poor quality if automated Content-based retrieval systems n 10/27/2020 Support retrieval based on the image content, such as color histogram, texture, shape, objects, and wavelet transforms Data Mining: Principles and Algorithms 68

Queries in Content-Based Retrieval Systems n Image sample-based queries n n n Find all of the images that are similar to the given image sample Compare the feature vector (signature) extracted from the sample with the feature vectors of images that have already been extracted and indexed in the image database Image feature specification queries n n 10/27/2020 Specify or sketch image features like color, texture, or shape, which are translated into a feature vector Match the feature vector with the feature vectors of the images in the database Data Mining: Principles and Algorithms 69

Approaches Based on Image Signature n Color histogram-based signature n n The signature includes color histograms based on color composition of an image regardless of its scale or orientation No information about shape, location, or texture Two images with similar color composition may contain very different shapes or textures, and thus could be completely unrelated in semantics Multifeature composed signature n 10/27/2020 Define different distance functions for color, shape, location, and texture, and subsequently combine them to derive the overall result Data Mining: Principles and Algorithms 70

Wavelet Analysis n Wavelet-based signature n Use the dominant wavelet coefficients of an image as its signature n Wavelets capture shape, texture, and location information in a single unified framework n Improved efficiency and reduced the need for providing multiple search primitives n May fail to identify images containing similar objects that are in different locations. 10/27/2020 Data Mining: Principles and Algorithms 71

One Signature for the Entire Image? n n n Walnus: [NRS 99] by Natsev, Rastogi, and Shim Similar images may contain similar regions, but a region in one image could be a translation or scaling of a matching region in the other Wavelet-based signature with region-based granularity n Define regions by clustering signatures of windows of varying sizes within the image n Signature of a region is the centroid of the cluster n Similarity is defined in terms of the fraction of the area of the two images covered by matching pairs of regions from two images 10/27/2020 Data Mining: Principles and Algorithms 72

Multidimensional Analysis of Multimedia Data n n Multimedia data cube n Design and construction similar to that of traditional data cubes from relational data n Contain additional dimensions and measures for multimedia information, such as color, texture, and shape The database does not store images but their descriptors n Feature descriptor: a set of vectors for each visual characteristic n n 10/27/2020 Color vector: contains the color histogram MFC (Most Frequent Color) vector: five color centroids MFO (Most Frequent Orientation) vector: five edge orientation centroids Layout descriptor: contains a color layout vector and an edge layout vector Data Mining: Principles and Algorithms 73

Multi-Dimensional Search in Multimedia Databases 10/27/2020 Data Mining: Principles and Algorithms 74

Multi-Dimensional Analysis in Multimedia Databases Color histogram 10/27/2020 Texture layout Data Mining: Principles and Algorithms 75

Mining Multimedia Databases Refining or combining searches Search for “airplane in blue sky” (top layout grid is blue and keyword = “airplane”) Search for “blue sky” (top layout grid is blue) 10/27/2020 Search for “blue sky and green meadows” (top layout grid is blue and bottom is green) Data Mining: Principles and Algorithms 76

Mining Multimedia Databases The Data Cube and the Sub-Space Measurements JP EG GI By Size F all Sm edium ge M arge y Lar L er V By Format & Size RED WHITE BLUE Cross Tab JPEG GIF By Colour RED WHITE BLUE Group By Colour RED WHITE BLUE Measurement 10/27/2020 Sum By Colour & Size Sum By Format & Colour By Colour • Format of image • Duration • Colors • Textures • Keywords • Size • Width • Height • Internet domain of image • Internet domain of parent pages • Image popularity Data Mining: Principles and Algorithms 77

Mining Multimedia Databases in 10/27/2020 Data Mining: Principles and Algorithms 78

Classification in Multi. Media. Miner 10/27/2020 Data Mining: Principles and Algorithms 79

Mining Associations in Multimedia Data n Special features: n Need # of occurrences besides Boolean existence, e. g. , n “Two red square and one blue circle” implies theme “air-show” n Need spatial relationships n Blue on top of white squared object is associated with brown bottom n Need multi-resolution and progressive refinement mining n It is expensive to explore detailed associations among objects at high resolution n It is crucial to ensure the completeness of search at multi-resolution space 10/27/2020 Data Mining: Principles and Algorithms 80

Mining Multimedia Databases Spatial Relationships from Layout property P 1 on-top-of property P 2 property P 1 next-to property P 2 Different Resolution Hierarchy 10/27/2020 Data Mining: Principles and Algorithms 81

Mining Multimedia Databases From Coarse to Fine Resolution Mining 10/27/2020 Data Mining: Principles and Algorithms 82

Challenge: Curse of Dimensionality n n Difficult to implement a data cube efficiently given a large number of dimensions, especially serious in the case of multimedia data cubes Many of these attributes are set-oriented instead of single -valued Restricting number of dimensions may lead to the modeling of an image at a rather rough, limited, and imprecise scale More research is needed to strike a balance between efficiency and power of representation 10/27/2020 Data Mining: Principles and Algorithms 83

Summary n n Mining object data needs feature/attribute-based generalization methods Spatial, spatiotemporal and multimedia data mining is one of important research frontiers in data mining with broad applications Spatial data warehousing, OLAP and mining facilitates multidimensional spatial analysis and finding spatial associations, classifications and trends Multimedia data mining needs content-based retrieval and similarity search integrated with mining methods 10/27/2020 Data Mining: Principles and Algorithms 84

References on Spatial Data Mining n n n n H. Miller and J. Han (eds. ), Geographic Data Mining and Knowledge Discovery, Taylor and Francis, 2001. Ester M. , Frommelt A. , Kriegel H. -P. , Sander J. : Spatial Data Mining: Database Primitives, Algorithms and Efficient DBMS Support, Data Mining and Knowledge Discovery, 4: 193 -216, 2000. J. Han, M. Kamber, and A. K. H. Tung, "Spatial Clustering Methods in Data Mining: A Survey", in H. Miller and J. Han (eds. ), Geographic Data Mining and Knowledge Discovery, Taylor and Francis, 2000. Y. Bedard, T. Merrett, and J. Han, "Fundamentals of Geospatial Data Warehousing for Geographic Knowledge Discovery", in H. Miller and J. Han (eds. ), Geographic Data Mining and Knowledge Discovery, Taylor and Francis, 2000 K. Koperski and J. Han. Discovery of spatial association rules in geographic information databases. SSD'95. Shashi Shekhar and Sanjay Chawla, Spatial Databases: A Tour , Prentice Hall, 2003 (ISBN 013 -017480 -7). Chapter 7. : Introduction to Spatial Data Mining X. Li, J. Han, and S. Kim, Motion-Alert: Automatic Anomaly Detection in Massive Moving Objects”, IEEE Int. Conf. on Intelligence and Security Informatics (ISI'06). 10/27/2020 Data Mining: Principles and Algorithms 85

10/27/2020 Data Mining: Principles and Algorithms 86