Outlier Analysis Second Edition - Charu Aggarwal

1 Outlier AnalysisSecond EditionCharu C. AggarwalIBM T. J. Watson Research CenterYorktown Heights, New YorkNovember 25, 2016 PDF Downloadable from iiPDF Downloadable from To my wife, my daughter Sayani,and my late parents Dr. Prem Sarup and Mrs. Pushplata Downloadable from ivPDF Downloadable from Contents1 An Introduction to Outlier TheDataModelisEverything .. Connections with Supervised The Basic Outlier Detection Models .. Feature Selection in Outlier Probabilistic and Statistical LinearModels .. Spectral High-Dimensional Outlier OutlierEnsembles.

2 SequentialEnsembles .. IndependentEnsembles .. Categorical,Text, Times-SeriesDataandDataStreams .. SpatialData .. Supervised Outlier InterpretingtheROCAUC .. CommonMistakesinBenchmarking .. 33vPDF Downloadable from viCONTENTS2 Probabilistic Models for Outlier Probabilistic Tail Inequalities .. Sum of Bounded Random Statistical-Tail Confidence .. Sum of Squares of Extreme-ValueAnalysisinMultivariateData .. Angle-Based Outlier Detection .. Distance Distribution-based Techniques: The Mahalanobis Method.

3 Probabilistic Mixture Modeling for Outlier RelationshipwithClusteringMethods .. OtherWaysofLeveragingtheEMModel .. An Application of EM for Converting Scores to Limitations of Probabilistic Modeling .. Exercises .. 623 Linear Models for Outlier LinearRegressionModels .. ModelingwithDependentVariables .. Linear Modeling with Mean-Squared Projection PrincipalComponentAnalysis .. Connections with the Mahalanobis Applications to Noise How Many Eigenvectors?.. ExtensiontoNonlinearDataDistributions.

4 ApplicationtoArbitraryDataTypes .. One-Class Support Vector Machines .. SolvingtheDualOptimizationProblem .. Connections to Support Vector Data Description and Other KernelModels .. AMatrixFactorizationViewofLinearModels .. 95 PDF Downloadable from Outlier Detection in Incomplete ComputingtheOutlierScores .. GeneralizationtoNonlinearModels .. LimitationsofLinearModeling .. 1094 Proximity-Based Outlier ApplicationtoArbitraryDataTypes .. AdvantagesandDisadvantagesofClusteringMe thods .. Distance-BasedOutlierAnalysis.

5 BinaryOutputsforDistance-BasedMethods .. Cell-BasedPruning .. Sampling-BasedPruning .. Index-BasedPruning .. IntensionalKnowledgeofDistance-BasedOutl iers .. Handling Duplicate Points and Stability LOCIPlot .. Histogram-BasedTechniques .. KernelDensityEstimation .. Connection with Harmonick-Nearest Neighbor Detector . LocalVariationsofKernelMethods .. Ensemble-Based Implementations of Histograms and Kernel Methods Limitations of Proximity-Based Exercises .. 1465 High-Dimensional Outlier Axis-Parallel Genetic Algorithms for Outlier Defining Abnormal Lower-Dimensional Defining Genetic Operators for Subspace 154 PDF Downloadable from Finding Distance-Based Outlying Feature Bagging: A Subspace Sampling Projected Clustering Ensembles.

6 Subspace Histograms in Linear IsolationForests .. Further Enhancements for Subspace Relationship to Clustering Ensembles and Histograms .. Selecting High-Contrast Local Selection of Subspace Distance-BasedReferenceSets .. Generalized Generalized Projected Clustering LeveragingInstance-SpecificReferenceSets .. Rotated Subspace Sampling .. Nonlinear RegressionModelingTechniques .. Discussion of Subspace Exercises .. 1846 Outlier CategorizationandDesignofEnsembleMethods .. BasicScoreNormalizationandCombinationMet hods.

7 Theoretical Foundations of Outlier What is the Expectation Computed Over?.. VarianceReductionMethods .. ParametricEnsembles .. Randomized Detector Feature Bagging: An Ensemble-Centric Connections to Representational Weaknesses of Feature Rotated IsolationForests:AnEnsemble-CentricView .. Data-CentricVarianceReductionwithSamplin g .. Variable Subsampling .. Variable Subsampling with Rotated Bagging (VR).. OtherVarianceReductionMethods .. Flying Blind with Bias BiasReductionbyData-CentricPruning .. 215 PDF Downloadable from Exercises.

8 2187 Supervised Outlier Full Supervision: Rare Class Detection .. Cost-SensitiveLearning .. WeightingMethods .. SyntheticOver- BoostingMethods .. Semi- Semi-Supervision:PartiallyObservedClasse s .. One-ClassLearningwithNormalExamples .. Learning with a Subset of Labeled Supervised Models for Unsupervised Outlier Connections with PCA-Based Applicability to Mixed-Attribute Data OtherClassificationMethodswithSyntheticO utliers .. 2478 Categorical, Text, and Mixed Attribute Extending Probabilistic Models to Categorical ModelingMixedData.

9 ExtendingLinearModelstoCategoricalandMix edData .. LeveragingSupervisedRegressionModels .. ContextualSimilarity .. Connections to Linear IssueswithMixedData .. Density-BasedMethods .. ClusteringMethods .. Outlier Detection in Binary and Transaction Subspace Outlier Detection in Text 262 PDF Downloadable from Probabilistic LinearModels:LatentSemanticAnalysis .. Probabilistic Latent Semantic Analysis (PLSA).. First Story Detection .. Exercises .. 2729 Time Series and Streaming Outlier Predictive Outlier Detection in Streaming Time-Series Selection Principal Component Analysis and Hidden Variable-BasedModels.

10 Relationship between Unsupervised Outlier Detection and Prediction Supervised Point Outlier Detection in Time Time-SeriesofUnusualShapes .. TransformationtoOtherRepresentations .. Leveraging Trajectory Representations of Time Series .. Probabilistic LinearModels .. LeveragingKernelMethodswithLinearModels .. SupervisedMethodsforFindingUnusualTime-S eriesShapes .. Multidimensional Streaming Outlier IndividualDataPointsasOutliers .. Proximity-BasedAlgorithms .. Probabilistic High-DimensionalScenario.