Tutorial on Spatial and Spatio-Temporal Data Mining Part II ... - UFSC

<strong>Tutorial</strong> on Spatial and Spatio-Temporal Data MiningPart II – Trajectory Knowledge DiscoveryVania BogornyUniversidade Federal de Santa Catarinawww.inf.ufsc.br/~vaniavania@inf.ufsc.brOutlineThe wireless explosionMoving Object Data and Mobility Data AnalysisTrajectory PatternsGeometric Trajectory Pattern Mining Methods:TT 32T 1T 4Semantic Trajectory Pattern Mining Methods:RCRCSCHT T 32 HT 4T 1Trajectory Data mining ToolsHHotelRRestaurant CCinema

The Wireless Explosion (Fosca Giannotti 2007 – www.geopkdd.eu)Have you ever feel to be tracked?The Wireless ExplosionThe world becomes more and more mobile with the easyaccess to smart phones, GPS, etcSattelite services, sensors and wireless technologies arerapidly improvinglots of spatio-temporal data is being generated

The Wireless A Explosão Explosion da Rede Sem Fio(Fosca Giannotti 2007 – www.geopkdd.eu)Mobile devices leave behind digital traces that are collected astrajectories, describing the movement of its usersMobile devices generate a new type of data, called “ Trajectories ofMoving Objects”5Mobility Data Analysis

Mobility Data AnalysisSeveral analysis may be done over trajectories:How people move around the townDuring the day, during the week, etc.Are there typical movement behaviours? In a certain area at a certaintime?How are people movement habits changing in this area in last decadeyear-month-day?Are there relations between movements of two areas?Are there periodic movements?Mobility Data Analysis: ApplicationsTrajectory data analysis may be useful inseveral application domainsVeicule MonitoringTransportation Companies monitor their trucksInsurance companies use GPS devices to monitorinsured vehicles to reduce insurance priceTraffic AnalysisTo alert people about traffic jams,accidents, etc...Identify/predict low traffic regionsin a city

Mobility Data AnalysisAnimal Migration / Behaviour AnalysisWhich are the trajectories ofa given migration bird?Where do birds stop? For how long?Which is the migration pattern of certain species?Fishing Analysis and ControlAre boats really fishing inallowed areas?Can we classify vessel trajectories?Mobility Data AnalysisWeather prediction and movementanalysisHurricane tracking

Trajectory DataTrajectory Data (Giannotti 2007 – www.geopkdd.eu)Spatio-temporal DataRepresented as a set of points, located in space and timeT=(x 1 ,y 1 , t 1 ), …, (x n , y n , t n ) => position in space at time t i was(x i ,y i )Tid position (x,y) time (t)1 48.890018 2.246100 08:251 48.890018 2.246100 08:26... ... ...1 48.890020 2.246102 08:401 48.888880 2.248208 08:411 48.885732 2.255031 08:42... ... ...1 48.858434 2.336105 09:041 48.853611 2.349190 09:05... ... ...2 ... ...

Trajectories: Overall Characteristics(Adrienko 2008)1. Geometric shape2. Length (traveled distance)3. Duration (in time)4. SpeedMean and maximal SpeedAcceleration, deceleration5. Direction:Periods of straight, curvilinear, circular movementMore.....RelationshipsMany types of relations may be of interest, depending on the problem:similarity or difference of the overall characteristics of the trajectoriese.g. shapes, travelled distances, durations, dynamics of speed and directions)spatial and temporal relations:co-location in space (i.e. the trajectories consist of the samepositions or have some positions in common):co-existence in time (i.e. the trajectories are collected during thesame time period or the periods overlap);co-incidence in space and time (i.e. same positions are attained atthe same time);distances in space and in time.

Trajectory PatternsMining Trajectories: ClusteringFosca Giannotti 2007 – www.geopkdd.euGroup together similar trajectoriesFor each group produce a summary= cell

Mining Trajectories : Frequent patternsFosca Giannotti 2007 – www.geopkdd.euFrequent followed paths= cellMining Trajectories: classification modelsFosca Giannotti 2007 – www.geopkdd.euExtract behaviour rules from historyUse them to predict behaviour of future users20%7%?5%60%8%= cell

Trajectory Data Mining MethodsSpatio-Temporal Data Mining MethodsTwo approaches:Geometry-based spatio-temporal data mining:Density-based clustering methodsFocus on physical similarityConsider only geometrical properties of trajectories (space andtime)Semantic-based spatio-temporal data miningDeal with sparse data alsoPatterns are computed based on the semantics of the dataTrajectories are pre-processed to enrich the data

Geometry-based Trajectory Data MiningMethodsGeneral Geometric Trajectory Patterns

Relative Motion Patterns (Laube 2004)Proposed 5 kinks of trajectory patterns based on movement,direction, and location: convergence, encounter, flock,leadership, and recurrenceConvergence: At least m entities pass through the samecircular region of radius r, not necessarily at the sametime (e.g. people moving to train station)T 1T 2T 3T 4T 5convergenceRelative Motion Patterns (Laube 2004)Flock pattern: At least m entities are within a region of radius r and move inthe same direction during a time interval >= s (e.g. traffic jam)Leadership: At least m entities are within a circular region of radius r, theymove in the same direction, and at least one of the entities is heading in thatdirection for at least t time steps. (e.g. bird migration, traffic accident)Encounter: At least m entities will be concurrently inside the same circularregion of radius r, assuming they move with the same speed and direction.(e.g. traffic jam at some moment if cars keep moving in the same direction)T 2T 3LeadershipT 1EncounterFlock

Relative Motion Patterns (Laube 2004)Recurrence: at least m entities visit acircular region at least k timesF1F 1RecurrenceF 1F1Extension of the work proposed by [Laube 2004, 2005]Gudmundsson(2006)Computes the longest duration flock patternsThe longest pattern has the longest durationAnd has at least a minimal number oftrajectoriesGudmundsson (2007)proposes approximate algorithms for computingthe patterns leadership, encounter,convergence, and flockFocus relies on performance issues

Frequent Trajectory PatternsFrequent Mobile Group Patterns (Hwang, 2005)A group pattern is a set of trajectories close to each other(with distance less than a given minDist) for a minimalamount of time (minTime)Direction is not consideredFrequent groups are computed with the algorithm AprioriGroup pattern: time, distance, and minsup

Co-Location Patterns (Cao 2006)Co-location episoids in spatio-temporal dataTrajectories are spatially close in a time window and move togetherw2w1Traclus (Han, 2007)Clustering algorithm (TraClus-Trajectory Clustering)Group sub-trajectoriesDensity-basedPartition-and-group method1) each trajectory is partitioned into a set of line segments (subtrajectories)with lenght L defined by the user2) similar segments (close segments) are groupedSimilarity is based on a distance functionInteresting approach for trajectories of hurricanesMain drawback: Clustering is based on spatial distancetime is not considerd

Trajectory Sequential PatternsFrequent Sequential Patterns (Cao, 2005)Three main steps:1. Transforms each trajectory in a line with several segments A distance tolerance measure is defined (similar to buffer) All trajectory points inside this distanceare summarized in one segment2. Similar segments are grouped Similarity is based on the angle and the spatial lenght of the segmentSegments with same angle and length havetheir distance checked based on a given distance d thresholdFrom the resultant groups, a medium segment is createdFrom this segment a region (buffer) is created3. Frequent sequences of regions are computedconsidering a minSup threshold

T-Patterns (Giannotti, 2007)Sequential Trajectory Pattern MiningConsider both space and timeObjective is to describe frequent movementConsidering visited regions of interestDuring movements and the duration of movementsSteps:1. Compute or find regions of interest, based on dense spatialregions (no time is considered)2. Select trajectories that intersect two or more regions in asequence, annotating travel time from one region to another3. Compute sequences of regions visited in same time intervalsT-Patterns (Giannotti, 2007)Fix a set of pre-defined regionsBACMap each (x,y) of the trajectory to its regiontimeSample pattern:A⎯ 20min.⎯⎯→B

T-Patterns (Giannotti, 2007)Detect significant regions thru spatial clusteringaround(x 1 ,y 1 )around(x 1 ,y 1 )Map each (x,y) of the trajectory to its regiontimeSample pattern:20 min.around ( x1,y1)⎯⎯⎯→ around ( x2,y2)Trajectory ClassificationThe idea is to classify types of trajectories

TraClass Algorithm (Lee 2008)Two main steps algorithm:First: region – based clustering:Second: trajectory-clusteringMain problem: time is not consideredTraClass Algorithm (Lee 2008)Classify subtrajectories instead of whole trajectoriesExamples: Red trajectories move from Port A to Container Port and then to Port B Blue trajectories move from Port A to Refinery and then to Port BClassifying whole trajectory would classify all trajectories as moving fromPort A to Port B38

TraClass Algorithm (Lee 2008)First: region – based clusteringTrajectories are cut into segments (fast change of direction)Segments are then clustered by distance with DB-SCANOne representative trajectory is generated for the cluster and labeled witha classTR TRTR 4 53 (1)A set of trajectories(2) PartitionTR 2TR 1A representative trajectoryA set of line segments(3) GroupA cluster39TraClass Algorithm (Lee 2008)First: Discover regions that have trajectories mostly of one class regardless oftheir movement patterns

TraClass Algorithm (Lee 2008)Second: trajectory – based clustering:Extracts clusters of common movementpatterns in non-homogeneous areasGrouping is based on same class41Trajectory Outlier Detection

Trajectory Outlier Detection• The objective is to find trajectories that have different behaviorin relation to other trajectories• For instance:– A fishing vessel that has a behaviour differentfrom other fishing vessels in the same area– A hurricane that may change behaviour in certainparts of its trajectory– Cars or pedestrians with suspishious behaviour43TraOD - Trajectory Outlier Detection (Lee 2008)• Partition trajectories into subtrajectories• Compare subtrajectories based on:– distance and length• If a subtrajectory is not close to othertrajectories for a minimal lenght– It is an outlier44

TraOD - Trajectory Outlier Detection (Lee 2008)• Example:– Looking to the whole trajectory, TR 3 is not detected asan outlier since its overall behavior is similar toneigbouhr trajectories• Looking at the subtrajectories, T3 can be an outlierTR 5TR 1TR 4 TR3TR2An outlying sub-trajectory45TraOD - Trajectory Outlier Detection (Lee 2008)Two phases: partitioning and detectionTR 5TR 4 TR3TRTR 21(1) PartitionA set of trajectoriesA set of trajectory partitions(2) DetectTR 3An outlierOutlying trajectory partitions46

TraOD - Trajectory Outlier Detection (Lee 2008)• Once trajectories are partitioned, trajectory outliers aredetected based on both distance and density• A trajectory is an outlier if it contains a sufficient amount ofoutlying t-partitionsNot closeClose ≤ 1‒p> 1‒pTRi LiTRi LiL i is an outlying t-partitionL i is not an outlying t-partition47TraOD - Trajectory Outlier Detection (Lee 2008)13 Outliers from Hurricane Data48

SummaryThese data mining approaches deal with Trajectory SamplesTid geometry timest1 48.890018 2.246100 08:251 48.890018 2.246100 08:26... ... ...1 48.890020 2.246102 08:401 48.888880 2.248208 08:411 48.885732 2.255031 08:42... ... ...1 48.858434 2.336105 09:041 48.853611 2.349190 09:05... ... ...1 48.853610 2.349205 09:401 48.860515 2.349018 09:41... ... ...1 48.861112 2.334167 10:001 48.861531 2.336018 10:011 48.861530 2.336020 10:02... ... ...2 ... ...ReferencesLaube, P. and Imfeld, S. (2002). Analyzing relative motion within groups of trackablemoving point objects. In Egenhofer, M. J. and Mark, D. M., editors, GIScience, volume2478 of Lecture Notes in Computer Science, pages 132–144. Springer.Laube, P., Imfeld, S., and Weibel, R. (2005a). Discovering relative motion patterns ingroups of moving point objects. International Journal of Geographical InformationScience, 19(6):639–668.Laube, P., van Kreveld, M., and Imfeld, S. (2005b). Finding REMO: Detecting RelativeMotion Patterns in Geospatial Lifelines. Springer.Lee, J.-G., Han, J., and Whang, K.-Y. (2007). Trajectory clustering: a partition-and-groupframework. In Chan, C. Y., Ooi, B. C., and Zhou, A., editors, SIGMOD Conference,pages 593–604. ACM.Li, Y., Han, J., and Yang, J. (2004). Clustering moving objects. In KDD ’04: Proceedings ofthe tenth ACM SIGKDD international conference on Knowledge discovery and datamining, pages 617–622, New York, NY, USA. ACM Press.Nanni, M. and Pedreschi, D. (2006). Time-focused clustering of trajectories of movingobjects. Journal of Intelligent Information Systems, 27(3):267–289.

ReferencesVerhein, F. and Chawla, S. (2006). Mining spatio-temporal association rules, sources, sinks,stationary regions and thoroughfares in object mobility databases. In Lee, M.- L., Tan,K.-L., and Wuwongse, V., editors, DASFAA, volume 3882 of Lecture Notes in ComputerScience, pages 187–201. Springer.Gudmundsson, J. and van Kreveld, M. J. (2006). Computing longest duration flocks intrajectory data. In [de By and Nittel 2006], pages 35–42.Gudmundsson, J., van Kreveld, M. J., and Speckmann, B. (2007). Efficient detection ofpatterns in 2d trajectories of moving points. GeoInformatica, 11(2):195–215.Hwang, S.-Y., Liu, Y.-H., Chiu, J.-K., and Lim, E.-P. (2005). Mining mobile group patterns: Atrajectory-based approach. In Ho, T. B., Cheung, D. W.-L., and Liu, H., editors, PAKDD,volume 3518 of Lecture Notes in Computer Science, pages 713–718. Springer.Cao, H., Mamoulis, N., and Cheung, D. W. (2006). Discovery of collocation episodes inspatiotemporal data. In ICDM, pages 823–827. IEEE Computer Society.Zhenhui Li, Jae-Gil Lee, Xiaolei Li, Jiawei Han: Incremental Clustering for Trajectories.DASFAA (2) 2010: 32-46More...Huiping Cao, Nikos Mamoulis, David W. Cheung: Discovery of Periodic Patterns in SpatiotemporalSequences. IEEE Trans. Knowl. Data Eng. 19(4): 453-467 (2007)Panos Kalnis, Nikos Mamoulis, Spiridon Bakiras: On Discovering Moving Clusters in SpatiotemporalData. SSTD, 364-381 (2005)Florian Verhein, Sanjay Chawla: Mining spatio-temporal patterns in object mobility databases.Data Min. Knowl. Discov. 16(1): 5-38 (2008)Florian Verhein, Sanjay Chawla: Mining Spatio-temporal Association Rules, Sources, Sinks,Stationary Regions and Thoroughfares in Object Mobility Databases. DASFAA, 187-201(2006)Cao, H., Mamoulis, N., and Cheung, D. W. (2005). Mining frequent spatio-temporal sequentialpatterns. In ICDM ’05: Proceedings of the Fifth IEEE International Conference on Data Mining,pages 82–89, Washington, DC, USA. IEEE Computer Society.Jae-Gil Lee, Jiawei Han, Xiaolei Li, and Hector Gonzalez, “TraClass: Trajectory ClassificationUsing Hierarchical Region-Based and Trajectory-Based Clustering”, Proc. 2008 Int. Conf.on Very Large Data Base (VLDB'08), Auckland, New Zealand, Aug. 2008.Jae-Gil Lee, Jiawei Han, and Xiaolei Li, "Trajectory Outlier Detection: A Partition-and-DetectFramework", Proc. 2008 Int. Conf. on Data Engineering (ICDE'08), Cancun, Mexico, April2008.

Semantic-based Spatio-temporal Data MiningMethodsSemantic Trajectory Data MiningThe main idea is to enrich trajectories with domainsemantic information in preprocessing stepsThis task can be done using data miningApply data mining as a second stepMining is on semantic rich trajectories

Geometric Patterns X Semantic Patterns (Bogorny 2008)Geometric PatternRTPRTPCCT 3T 2T 1HCCT 2T 3T 4HT 4T 1HHotelRRestaurantTP TouristicPlaceSemantic trajectory Pattern(a) Hotel to Restaurant, passing by CC(b) go to Cinema, passing by CCGeometric Patterns X Semantic Patterns (Bogorny 2008)There is very little or no semantics in most DM approaches fortrajectoriesConsequence:• Patterns are purely geometrical• Difficult to interpret from the user’s point of view• Do not discover semantic patterns,which can be independent of spatial location

DJ-Cluster (Zhou 2007)DJ-Cluster is a variation of DBSCANFocus relies on performance issuesObjective: find interesting places of individual trajectoriesClusters are computed from a SET of trajectories of the same objectTime is not consideredA Conceptual View on Trajectories (Spaccapietra 2008)A trajectory is a spatio-temporal thing (an object) thathas generic featuresgeneric: application independenthas semantic featuressemantic: application dependentA trajectory is more than a moving object

Semantic Trajectories - MotivationTrajectory Samples (x,y,t)Geographic DataGeographic Data +Trajectory Data =Semantic TrajectoriesThe Model of Stops and Moves (Spaccapietra 2008)STOPSImportant parts of trajectoriesWhere the moving object has stayed for aminimal amount of timeStops are application dependentTourism application– Hotels, touristic places, airport, Traffic Management Application– Traffic lights, roundabouts, big eventsMOVESAre the parts that are not stops

Semantic TrajectoriesA semantic trajectory is a set of stops and movesStops have a place, a start time and an end timeMoves are characterized by two consecutive stopsMethods for Adding Semantics to TrajectoriesPre-processing Single Trajectories

Methods to Compute Stops and Moves1) IB-SMoT (INTERSECTION-based)Interesting for applications like tourism and urban planning2) CB-SMoT (SPEED-based clustering)Interesting for applications where the speed is important,like traffic management3) DB-SMOT (DIRECTION-based clustering)Interesting in application where the direction variation is importantlike fishing activitiesIB-SMoT (Alvares 2007a)A candidate stop C is a tuple (R C , ∆ C ), whereR C is the geometry of the candidate stop (spatial feature type)∆ C is the minimal time durationE.g. [Hotel - 3 hours]An application A is a finite setA = {C 1 = (R C1 , ∆ C1 ), …, C N = (R CN , ∆ CN )} of candidatestops with non-overlapping geometries R C1 , … ,R CNE.g. [Hotel - 3 hours, Museum – 1 hour]

IB-SMoT (Alvares 2007a)A stop of a trajectory T is a place that is important for theapplicationA move of T with respect to an application is: a maximal contiguous subtrajectory of T :between the starting point of T and the first stop of T; ORbetween two consecutive stops of T; ORS1between the last stop of T and the ending point of T; or the trajectory T itself, if T has no stops.S2S3IB-SMoT(Alvares 2007ª)Input: candidate stopstrajectoriesOutput: Semantic rich trajectories// Application// trajectory samplesMethod:For each trajectoryCheck if it intersects a candidat stop for a minimal amount of timeJurere09-12IbisH.13-14FloripaS16-17

CB-SMoT: Speed-based clustering (Palma 2008)• Clusters single trajectories based on the speed variation:low speed important placeCB-SMoT: Speed-based clustering (Palma 2008)Jurere09-12Unknown stopInput: Trajectory samplesSpeed variationminTimeOutput: stops and movesStep 1: find clustersStep 2: Add semantics to eachclusterIbisH.13-14FloripaS16-172.1: If intersects α during ∆tα stop α2.2: If no intersectionduring ∆t unknown stop

CB-SMoT: Speed-based clustering (Palma 2008)Unknown Stops (CB-SMOT)T 1same unknown stopT 2another unknown stopCB-SMoT: Speed-based clustering (Palma 2008)Can Find Clusters Inside Buildingsp 1p 6p 7p 11t6= 10:10AMt7= 10:32AM

DB-SMOT : Direction-based Clustering (Manso 2010)Input: trajectories // trajectory samplesminDirVariation // minimal direction variationminTime // minimum timemaxToleranceOutput: semantic rich trajectoriesMethod:For each trajectoryFind clusters with direction variationhigher than minDirVariationFor a minimal amount of timeExamples of semantic trajectory patterns

Semantic Rich Trajectories (Transportation Application)IB-SMoTCB-SMoTSequential Patterns (Transportation Application)

Fishing DomainDB-SMoT MethodMultiple-granularity semantic trajectorypattern mining

STOPS at Multiple-Granularities (Bogorny 2009)Stop at Ibis Hotel from 6:04PM to 7:42PM, september 16, 2010spacetimeIbisHotel or Hotel or AccommodationAfternoon or Thursday or 6:00PM – 8:00PM or RUSH-HOURITEMS - the building blocks for semantic pattern discoveryAn item is generated either from a stop or a moveAn item is a set of complex information (space +time), that can be defined in many formats/typesand at different granularities

Building an ITEM for Data Mining (Bogorny 2009)Formats/types for an item:ameOnly: is the name of the stop/moveSTOPS: name of the spatial feature instance• IbisHotelMOVES: name of the two stops which define the move• SydneyAirport – IbisHotelameStart: is the name of the stop/move + start timeIbisHotel [morning]--stopLouvreMuseum [weekend]--stopIbisHotel-SydneyAirport [10:00AM-11:00AM] --move10/11/2010 GIScience 2010 – A conceptual data model for trajectory data miningVania Bogorny, Universidade Federal de Santa Catarina, Brazil, www.inf.ufsc.br/~vania79Building an ITEM for Data Mining (Bogorny 2009)ameEnd: name of a stop/move + end timeIbisHotel[morning]stopIbisHotel-SydneyAirport[10:00AM-11:00AM] moveameStartEnd: name of a stop/move + start time + end timeIbisHotel[08:00AM-11:00AM][1:00pm-6:00pm] stopLouvreMuseum[morning][afternoon] stopSydenyAirport– IbisHotel [10:00AM-11:00PM] [10:00AM-6:00PM]10/11/2010 GIScience 2010 – A conceptual data model for trajectory data miningVania Bogorny, Universidade Federal de Santa Catarina, Brazil, www.inf.ufsc.br/~vania80

Multiple-Granularity Semantic Trajectory DMQL (Bogorny 2009)ST-DMQL is an approach to semantically enrichtrajectories with domain informationAutormatically tranforms these semantic information intodifferent space and time granularitiesExtracts frequent patterns, association rules andsequential patterns from semantic trajectoriesMultiple Level Semantic Sequential PatternsLarge Sequences of Length 2 (ITEM=SPACE+Start_Time)(41803_street_5, 41803_street_5) Support: 7(41803_street_4, 41803_street_4) Support: 9(41803_street_4, 66655_street_4) Support: 5(41803_street_2, 41803_street_2) Support: 6(41803_street_8, 41803_street_8) Support: 5(41803_street_3, 0_unknown_3) Support: 5gidtime unit = monthSpatial feature type (stop name)

Multiple Level Semantic Sequential PatternsLarge Sequences of Length 2 (ITEM=SPACE+Start_Time)(41803_street_tuesday,41803_street_tuesday) Support: 9(41803_street_tuesday,66655_street_tuesday) Support: 5(41803_street_monday,66655_street_monday) Support: 5(41803_street_monday,41803_street_monday) Support: 11(41803_street_monday,0_unknown_monday) Support: 5(41803_street_thursday,41803_street_thursday) Support: 13(41803_street_thursday,0_unknown_thursday) Support: 6(41803_street_wednesday,41803_street_wednesday) Support: 7gidTime unit = Day of the weekSpatial feature type (stop name)WEKA-STPMThe previous semantic pattern mining approaches areimplemented in WEKA

Weka-STDMWeka-STDM

Current Works on Weka-STPMTrajectory VisualizationTrajectory CleaningNew methods for trajectory pre-processingTrajectory Behaviour PatternsRecent works have emerged on mining behaviourpatterns from trajectories

Athena (Baglioni 2009)Semantic-rich movement analysisWhich are the homeworktrajectories? Andthe common behaviorsof them?To answer these questionswe need to define what is ahome-work trajectory (orpattern)The concept of the homeworktrajectory can beencoded in a formalframework to automaticallyinfer which trajectories arehome-workAthena (Baglioni 2009)Supports the post processing / deductive phase ofthe KDD processBased on ontologies to represent domain knowledgeand to infer the semantic types of thepatterns/trajectories.Semantic classification of patters/trajectories indomain concepts based on the semanticcharacteristics

Athena (Baglioni 2009)1. ExampleStopSemanticTrajectoryMoveCommuterTrajectoryontologyCommuter trajectory≡ a trajectory frequentlystarting outside the city, stopping inside the city for a longtime and going back outside the cityAthena (Baglioni 2009)SELECT t.id, t.objectFROM Milano_trWHERE ‘Commuter’ inSEMANTIC(t.object)Given the trajectories, we query the system toidentify the ones whose type is commuter, i.e.satisfying the ontology definition

Pattern Interpretation (Rebeca 2010)This work focuses on postprocessing,trying to interpret thepatternsConsidering that the movementcontext is essential to correctlyinterpret and understand thepatternsCONTEXT = geography + thematicattributesPattern Interpretation (Rebeca 2010)1. Mining movement patterns (stops)2. Semantic enrichment: annotates patterns withinformation obtained from trajectory types defined inONTOLOGIES (e.g. the trajectory of a commuter)3. Use the enriched representation to automaticallyclassify patterns

Works Summarized in this <strong>Tutorial</strong>Geometric PatternMining Methods(mining is on samplepoints)Laube 2004, 2005Hwang 2005Gudmundson 2006, 2007Giannotti 2007Lee 2007Cao 2006, 2007Lee 2007, 2008a, 2008bLi 2010Semantic Pattern MiningMethods (GenerateSemantic Trajectories usingDM - mining is on SemanticTrajectories)Alvares 2007Zhou 2007Palma 2008Bogorny 2009Bogorny 2010Manso 2010Alvares 2010Behaviour PatternMining andInterpretation MethodsGiannotti 2009Baglioni 2009Rebeca 2010ReferencesBogorny, V. ; Bart Kuijpers, Luis Otávio Alvares: ST-DMQL: A Semantic TrajectoryData Mining Query Language. International Journal of Geographical InformationScience 23(10): 1245-1276 (2009)Palma, A. T; Bogorny, V.; Kuijpers, B.; Alvares, L.O. A Clustering-based Approachfor Discovering Interesting Places in Trajectories. In: 23rd Annual Symposium onApplied Computing, (ACM-SAC'08), Fortaleza, Ceara, 16-20 March (2008) Brazil.pp. 863-868.Spaccapietra, S., Parent, C., Damiani, M. L., de Macedo, J. A., Porto, F., andVangenot, C. (2008). A conceptual view on trajectories. Data and KnowledgeEngineering, 65(1):126–146.Alvares, L. O., Bogorny, V., Kuijpers, B., de Macedo, J. A. F., Moelans, B., andVaisman, A. (2007b). A model for enriching trajectories with semantic geographicalinformation. In ACM-GIS, pages 162–169, New York, NY, USA. ACM Press.Rebecca Ong, Monica Wachowicz, Mirco anni, Chiara Renso, From PatternDiscovery to Pattern Interpretation in Movement Data - IEEE SADM 2010

ReferencesManso, J. A. ; TIMES, V. C. ; Oliveira, G. ; ALVARES, L. O. ; BOGORNY, V. . DB-SMoT: A Direction-Based Spatio-Temporal Clustering Method. In: IEEEInternational Conference on Intelligent Systems (IS), 2010, Londres. Proceedings ofthe IEEE International Conference on Intelligent Systems, 2010. p. 114-119. 3.Alvares, Luis O. ; PALMA, Andrey ; Oliveira, G. ; BOGORNY, V. . Weka-STPM:from trajectory samples to semantic trajectories. In: Workshop de Sofware Livre,2010, Porto Alegre. WSL, 2010.Zhou, C.; Nupur Bhatnagar, Shashi Shekhar, Loren G. Terveen: Mining PersonallyImportant Places from GPS Tracks. ICDE Workshops 2007: 517-526Bogorny, V. ; Carlos Alberto Heuser, Luis Otávio Alvares: A Conceptual Data Modelfor Trajectory Data Mining. GIScience 2010: 1-15Miriam Baglioni, José Antônio Fernandes de Macêdo, Chiara Renso, RobertoTrasarti, Monica Wachowicz: Towards Semantic Interpretation of MovementBehavior. AGILE Conf. 2009: 271-288Fosca Giannotti, Mirco Nanni, Dino Pedreschi, Chiara Renso, RobertoTrasarti: Mining Mobility Behavior from Trajectory Data. CSE (4) 2009:948-951Summary, Challenges and Open Issues in Spatio-Temporal Data Mining

Challenges and Open Issues in Spatio-Temporal Data MiningTrajectory ClusteringMost works are density-based clustering methodsMost are adapted spatial or non-spatial clustering algorithmsConsider either time or space, only a few consider both dimensionsChallenges and Open Issues in Spatio-Temporal Data MiningTrajectory SimilarityFocus relies on objective similarity measuresShape, direction, closenessNeeds: semantic similarityHigher abstraction level similarityExample:– groups of trajectories going together for shopping– Groups of trajectories going together to the University two timesa week

Challenges and Open Issues in Spatio-Temporal Data MiningNeed for data mining methods using:MetadataDomain knowledgeSemanticsOntologiesFor:Trajectory data pre-processingPattern pruningImprove the quality of the patternsPattern interpretationMore needsThere is a need for collaboration between data miners anddomain experts (environmental experts, transportationmanagers, metheorologists, etc)to evaluate data mining methods and the discovered patternsPost-Processing: almost no spatial or spatio-temporal datamining methods evaluate the patterns and theirinterestingness

Thank You !vania@inf.ufsc.br