Introduction to Machine Learning Techniques

ML Techniques

1. Course
Outcomes
After completion of this course, students will be able to
Understand machine-learning concepts.
Understand and implement Classification concepts.
Understand and analysethe different Regression
algorithms.
Apply the concept of Unsupervised Learning.
Apply the concepts of Artificial Neural Networks.

2. Topics
Introduction to ML:
Motivation and Applications
Importance of Data Visualization
Basics of Supervised, Unsupervised, and Reinforcement
Learning
Current research trends in ML

3. Machine
Learning
Introduction
MLisaninterdisciplinaryfield:
DataAnalyst:visualize,analyzedata,optimization
DataEngineers:buildandtestscalable/stable/
optimalecosystemsfordatascientiststoruntheir
algorithms
DatabaseAdministrator:responsibleforthe
properfunctioningofallthedatabases.
DataScientist:performpredictiveanalysisand
offeractionableinsights.
Statistician:extractandoffervaluableinsights
fromthedatausingstatisticaltheoryandtools.

4. Machine
Learning
Introduction

5. Machine
Learning
Introduction
AIstandsforArtificialIntelligence,andisbasicallythe
study/processwhichenablesmachinestomimichuman
behaviorthroughparticularalgorithm.
MLstandsforMachineLearning,andisthestudythatuses
statisticalmethodsenablingmachinestoimprovewith
experience.
DLstandsforDeepLearning,andisthestudythatmakesuse
ofNeuralNetworks(similartoneuronspresentinhumanbrain)
toimitatefunctionalityjustlikeahumanbrain.
Datascienceisthefieldofapplyingadvancedanalytics
techniquesandscientificprinciplestoextractvaluable
informationfromdataforbusinessdecision-making,strategic
planningandotheruses.

6. Evaluation of
Machine
Learning

7. Evaluation of
Machine
Learning
Continued

8. Evaluation of
Machine
Learning
Continued

9. What is
Human
Learning?
In cognitive science, learning is typically referred to as
the process of gaining information through
observation.
A task can be as simple as walking down the street or
doing the homework; or as complex as deciding the
angle in which a rocket should be launched so that it
can have a particular trajectory.
Why do we need to learn?
With more knowledge, the ability to do homework
with less number of mistakes increases
Thus, With more learning, tasks can be performed
more efficiently.

10. Types of
Human
Learning
1. Learning under expert guidance
Somebody who is an expert in the subject directly teaches us.
The process of gaining information from a person having
sufficient knowledge due to past experience. (e.g. learning of
child)
2. Learning guided by knowledge gained from experts
we build our own notion indirectly based on what we have
learnt from the expert in the past
learning also happens with the knowledge which has been
imparted by teacher or mentor at some point of time in some
other form
E.g. a kid can select one odd word from a set of words because
it is a verb and other words being all nouns, due to English
learned in school

11. Types of
Human
Learning
3. Learning by self
We do it ourselves, may be after multiple attempts,
some being unsuccessful.
Learning from our mistakes in past.
E.g. Child learning to walk through obstacles.

12. What is
Machine
Learning?
“Machinelearningisthefieldofstudythatgives
computerstheabilitytolearnwithoutbeing
explicitlyprogrammed”
-ArthurSamuel,AIpioneer,1959
“AcomputerprogramissaidtolearnfromexperienceE
withrespecttosomeclassoftasksTandperformance
measureP,ifitsperformanceattasksinT,asmeasuredby
P, improves with experience E”
-TomMitchell,MLProfessoratCMU
Algorithmsthat
improvetheirperformance(P)
atsometask(T)
withexperience(E)

13. Traditional v/s
Machine
Learning

14. How do
machine learn?
Data Input: Past data or information is utilized as a
basis for future decision-making
Abstraction: The input data is represented in a broader
way through the underlying algorithm
Generalization: The abstracted representation is
generalized to form a framework for making decisions

15. Well-posed
Learning
Problem
Fordefininganewproblem,whichcanbesolvedusingML,a
simpleframeworkcanbeused.Theframeworkinvolves
answeringthreequestions:
Whatistheproblem?
Describetheprobleminformallyandformallyandlist
assumptionsandsimilarproblems.
Whydoestheproblemneedtobesolved?
Listthemotivationforsolvingtheproblem,thebenefitsthatthe
solutionwillprovideandhowthesolutionwillbeused.
HowwouldIsolvetheproblem?
Describehowtheproblemwouldbesolvedmanuallytoflush
domainknowledge.

16. Machine
learning Life
cycle

17. Machine
learning Life
cycle
Machinelearninglifecycleinvolvessevenmajorsteps,which
aregivenbelow:
GatheringData
Datapreparation
DataWrangling
AnalyseData
Trainthemodel
Testthemodel
Deployment

18. 1. Gathering
Data
Data Gathering is the first step of the machine learning life cycle. The goal of
this step is to identify and obtain all data-related problems.
In this step, we need to identify the different data sources, as data can be
collected from various sources such asfiles,database,internet, ormobile
devices. It is one of the most important steps of the life cycle. The quantity
and quality of the collected data will determine the efficiency of the output.
The more will be the data, the more accurate will be the prediction.
This step includes the below tasks:
Identify various data sources
Collect data
Integrate the data obtained from different sources
By performing the above task, we get a coherent set of data, also called as
adataset. It will be used in further steps.

19. 2.Data
preparation
Aftercollectingthedata,weneedtoprepareitforfurthersteps.
Datapreparationisastepwhereweputourdataintoasuitable
placeandprepareittouseinourmachinelearningtraining.
Inthisstep,first,weputalldatatogether,andthenrandomizethe
orderingofdata.
Dataexploration:Itisusedtounderstandthenatureofdatathat
wehavetoworkwith.Weneedtounderstandthecharacteristics,
format,andqualityofdata.
A better understanding of data leads to an effective outcome. In
this, we find Correlations, general trends, and outliers.

20. 3. Data
Wrangling /
Data pre-
processing
Datawranglingistheprocessofcleaningandconvertingrawdataintoauseableformat.
Itistheprocessofcleaningthedata,selectingthevariabletouse,andtransformingthe
datainaproperformattomakeitmoresuitableforanalysisinthenextstep.Itisoneof
themostimportantstepsofthecompleteprocess.Cleaningofdataisrequiredto
addressthequalityissues.
Itisnotnecessarythatdatawehavecollectedisalwaysofouruseassomeofthedata
maynotbeuseful.Inreal-worldapplications,collecteddatamayhavevariousissues,
including:
MissingValues
Duplicatedata
Invaliddata
Noise
So,weusevariousfilteringtechniquestocleanthedata.
Itismandatorytodetectandremovetheaboveissuesbecauseitcannegativelyaffect
thequalityoftheoutcome.

21. 4. Data
Analysis
Nowthecleanedandprepareddataispassedontotheanalysis
step.Thisstepinvolves:
Selectionofanalyticaltechniques
Buildingmodels
Reviewtheresult
Theaimofthisstepistobuildamachinelearningmodelto
analyzethedatausingvariousanalyticaltechniquesandreview
theoutcome.Itstartswiththedeterminationofthetypeofthe
problems,whereweselectthemachinelearningtechniquessuch
asClassification,Regression,Clusteranalysis,Association,etc.
thenbuildthemodelusingprepareddata,andevaluatethe
model.
Hence,inthisstep,wetakethedataandusemachinelearning
algorithmstobuildthemodel.

22. 5. Train Model
Nowthenextstepistotrainthemodel,inthisstepwe
trainourmodeltoimproveitsperformanceforbetter
outcomeoftheproblem.
Weusedatasetstotrainthemodelusingvarious
machinelearningalgorithms.Trainingamodelis
requiredsothatitcanunderstandthevariouspatterns,
rules,and,features.

23. 6. Test Model
Onceourmachinelearningmodelhasbeentrainedon
agivendataset,thenwetestthemodel.Inthisstep,
wecheckfortheaccuracyofourmodelbyprovidinga
testdatasettoit.
Testingthemodeldeterminesthepercentageaccuracy
ofthemodelaspertherequirementofprojector
problem.

24. 7. Deployment
Thelaststepofmachinelearninglifecycleis
deployment,wherewedeploythemodelinthereal-
worldsystem.
Iftheabove-preparedmodelisproducinganaccurate
resultasperourrequirementwithacceptablespeed,
thenwedeploythemodelintherealsystem.But
beforedeployingtheproject,wewillcheckwhetherit
isimprovingitsperformanceusingavailabledataor
not.Thedeploymentphaseissimilartomakingthe
finalreportforaproject

25. Types of
Machine
Learning

26. Supervised
Learning

27. Supervised
Learning
Supervisedlearningisthetypesofmachinelearningin
whichmachinesaretrainedusingwell"labelled"
trainingdata,andonbasisofthatdata,machines
predicttheoutput.
Thelabelleddatameanssomeinputdataisalready
taggedwiththecorrectoutput.

28. Types of
Supervised
Learning
Classification (Discrete value output)Regression (Predict real value
output)

29. Unsupervised
Learning
Unsupervisedlearningisamachinelearning
techniqueinwhichmodelsarenotsupervisedusing
trainingdataset.
Instead,modelsitselffindthehiddenpatternsand
insightsfromthegivendata.Itcanbecomparedto
learningwhichtakesplaceinthehumanbrainwhile
learningnewthings.

30. Types of
Unsupervised
Learning
Clustering Association

31. Reinforcement
Learning
ReinforcementLearningisafeedback-based(reward)
Machinelearningtechniqueinwhichanagentlearnsto
behaveinanenvironmentbyperformingtheactions
andseeingtheresultsofactions.
Foreachgoodaction,theagentgetspositivefeedback,
andforeachbadaction,theagentgetsnegative
feedbackorpenalty.

32. Comparison –
Supervised,
Unsupervised
and
Reinforcement
Learning
CriteriaSupervised ML Unsupervised MLReinforcement ML
Definition
Learns by using
labelleddata
Trained using
unlabelleddata
without any
guidance.
Works on
interacting with the
environment
(rewardbased)
Type of dataLabelled data Unlabelled data
No –predefined
data
Type of
problems
Regression and
classification
Association and
Clustering
Exploitation or
Exploration
SupervisionExtra supervisionNo supervision No supervision
Algorithms
Linear Regression,
Logistic Regression,
SVM, KNN, NB, DT.
K –Means,
PCA, DBSCAN,
Apriori
Q –Learning,
SARSA
Aim Calculate outcomes
Discover underlying
patterns
Learn a series of
action
Application
Risk Evaluation,
Forecast Sales
Recommendation
System, Anomaly
Detection
Self Driving Cars,
Gaming, Healthcare

33. Did you know?
Manyvideogamesarebasedonartificialintelligence
techniquecalledExpertSystem.Thistechniquecan
imitateareasofhumanbehavior,withagoaltomimicthe
humanabilityofsenses,perception,andreasoning.

34. When not to
use ML?
Machinelearningshouldnotbeappliedtotasksin
whichhumansareveryeffectiveorfrequenthuman
interventionisneeded.
Forexample,airtrafficcontrolisaverycomplextask
needingintensehumaninvolvement.
Also,forverysimpletaskswhichcanbeimplemented
usingtraditionalprogrammingparadigms,thereisno
senseofusingmachinelearning.
Forexample,simplerule-drivenorformula-based
applicationslikepricecalculatorengine,dispute
trackingapplication,etc.donotneedmachinelearning
techniques.

35. Application of
ML

36. Tools for
Machine
Learning

37. Data
Visualization in
Machine
Learning
Datavisualizationisacrucialaspectofmachinelearningthat
enablesanalyststounderstandandmakesenseofdatapatterns,
relationships,andtrends.
Throughdatavisualization,insightsandpatternsindatacanbe
easilyinterpretedandcommunicatedtoawideraudience,making
itacriticalcomponentofmachinelearning.
Datavisualizationisthegraphicalrepresentationofinformation
anddata.
Byusingvisualelementslikecharts,graphs,andmaps,data
visualizationtoolsprovideanaccessiblewaytoseeand
understandtrends,outliers,andpatternsindata.

38. What is Data
Visualization?
Datavisualizationtranslatescomplexdatasets
intovisualformatsthatareeasierforthehumanbrain
tocomprehend.Thiscanincludeavarietyofvisual
toolssuchas:
Charts:Barcharts,linecharts,piecharts,etc.
Graphs:Scatterplots,histograms,etc.
Maps:Geographicmaps,heatmaps,etc.
Dashboards:Interactiveplatformsthatcombine
multiplevisualizations.

39. Types of Data
for
Visualization
Performingaccuratevisualizationofdataisverycritical
tomarketresearchwherebothnumericaland
categoricaldatacanbevisualized,whichhelpsincrease
theimpactofinsightsandalsohelpsinreducingthe
riskofanalysisparalysis.So,datavisualizationis
categorizedintothefollowingcategories:
NumericalData
CategoricalData

40. Types of Data
for
Visualization

41. Types of Data
Visualization
Approaches
Machine learning may make use of a wide variety of data
visualization approaches. That include:
Line Charts
Scatter Plots
Bar Charts
Heat Maps
Tree Maps
Box Plots

42. 1. Line Charts
Inalinechart,eachdatapointisrepresentedbyapoint
onthegraph,andthesepointsareconnectedbyaline.
Wemayfindpatternsandtrendsinthedataacross
timebyusinglinecharts.Time-seriesdataisfrequently
displayedusinglinecharts.

43. 2. Scatter Plots
Aquickandefficientmethodofdisplayingthe
relationshipbetweentwovariablesistousescatter
plots.Withonevariableplottedonthex-axisandthe
othervariabledrawnonthey-axis,eachdatapointina
scatterplotisrepresentedbyapointonthegraph.We
mayusescatterplotstovisualizedatatofindpatterns,
clusters,andoutliers.

44. 3. Bar Charts
Barchartsareacommonwayofdisplayingcategorical
data.Inabarchart,eachcategoryisrepresentedbya
bar,withtheheightofthebarindicatingthefrequency
orproportionofthatcategoryinthedata.Bargraphs
areusefulforcomparingseveralcategoriesandseeing
patternsovertime.

45. 4. Heat Maps
Heatmapsareatypeofgraphicalrepresentationthat
displaysdatainamatrixformat.Thevalueofthedata
pointthateachmatrixcellrepresentsdeterminesits
hue.Heatmapsareoftenusedtovisualizethe
correlationbetweenvariablesortoidentifypatternsin
time-seriesdata.

46. 5. Tree Maps
Treemapsareusedtodisplay
hierarchicaldatainacompact
formatandareusefulin
showingtherelationship
betweendifferentlevelsofa
hierarchy.

47. 6. Box Plots
Boxplotsareagraphicalrepresentationofthe
distributionofasetofdata.Inaboxplot,themedianis
shownbyalineinsidethebox,whilethecenterbox
depictstherangeofthedata.Thewhiskersextend
fromtheboxtothehighestandlowestvaluesinthe
data,excludingoutliers.Boxplotscanhelpusto
identifythespreadandskewnessofthedata.

48. Uses of Data
Visualization in
Machine
Learning
Identifytrendsandpatternsindata:Itmaybechallengingto
spottrendsandpatternsindatausingconventionalapproaches,
butdatavisualizationtoolsmaybeutilizedtodoso.
Communicateinsightstostakeholders:Datavisualizationcanbe
usedtocommunicateinsightstostakeholdersinaformatthatis
easilyunderstandableandcanhelptosupportdecision-making
processes.
Monitormachinelearningmodels:Datavisualizationcanbeused
tomonitormachinelearningmodelsinrealtimeandtoidentify
anyissuesoranomaliesinthedata.
Improvedataquality:Datavisualizationcanbeusedtoidentify
outliersandinconsistenciesinthedataandtoimprovedata
qualitybyremovingthem.