Chapter 12

Question 1

outlier detection, anaomaly detection

Answer 1

The process of finding data objecrts with nehaviors that ar very differne from expectation
such objects are called outliers or anomalies

Question 2

normal or expected data

Answer 2

data objexts that are not outliers which are abnormal data

Question 3

outlier

Answer 3

a data object tjhat deviates significantly from the normal objects as if it were genrated by a differnet mechanism

Question 4

nosie data

Answer 4

is random error or variance in a measured variable

noise should be removed before outlier detection

Question 5

three types of outliers

Answer 5

global, contextual and collective outlires

Question 6

global outlier or poit anaomaly

Answer 6

eobject is Og if it siginicatly deciates fiorm the rest of the data set
issue: find an appropriate measurement of deviation

Question 7

contexutal outlier or conditional outlier

Answer 7

object isOc if it deviates ignifcantly based on a selected context
ex. 80 degrees in urban, outlier depends on if it is winter oe summer
contextual attributes and behavioral attributes
generalization of local outlires
issue: how to define or formualre meaningful context

Question 8

contexual attibutes

Answer 8

deines the context, time or location

Question 9

behavioral attributes

Answer 9

chartacteristics of the object, used in outlier evaluation, ex temperature

Question 10

collective outliers

Answer 10

a subset of data objects collectively deciare significantly form the whole data ser even if the individual data objects may not be outliers
a data set may have multiple types of outliers
one object may belong to more than one type of outlier

Question 11

two outlier detection methods

Answer 11

based on whether user labeled examples of outliers can be obtained: supervised, semi supervised unserpervised methods
based on assumptions about normal data and outliers: statistical proximity based and cluseting based methods

Question 12

supervised methods

Answer 12

a classigfication problem
model normal objects ad report those not matching as otliers or treat those not matching the model as normal
ChallengesL imbalanced classes, catch as many outliers as possible, recall is more important than accuracy

Question 13

unserpervide methods

Answer 13

assume the normal objects aresomwaht clustered into multiple groups
an outlier is expected to be far way
weakneees: can not detect collective outliers effectively

Question 14

semi supervised methods

Answer 14

regarded as appluications of semi supervised learning
use labeled exampkse and the proximate unlabled object to train a model for normal objects
those not fitting the model are outliers

Question 15

statistical methods, model based methosa

Answer 15

assume that the normal data follow some statistical model, those that do not follow are outliers

Question 16

proximatiy based methods

Answer 16

an object is an outlier if the nearest neighbor of the object are far way, the proximity of the object is significantly deviates form the proximity of most of the other objects
distance based and density based