EssentialGuide toSeaborn
Visualstorytellingmadesimple.
Masterstatisticaldatavisualizationwithseaborn.Createinformative andbeautifulplotsdirectlyfromyourdata
EssentialGuideto Seaborn
IbonMartínez-Arranz
EssentialGuidetoSeaborn
IntroductionandSetup
WhySeaborn?
SeabornisaPythondatavisualizationlibrarybasedonMatplotlib.Itprovides ahigh-level,user-friendlyAPIforcreatinginformativeandattractivestatistical graphics.WhileMatplotlibishighlyflexibleandpowerful,Seaborno ersasetof toolsthatmakecommonvisualizationtasksfasterandeasier,especiallywhen workingwithpandasDataFrames.
WhyshouldyouuseSeaborn?
• EasyintegrationwithpandasDataFrames.
• Beautifuldefaultstyles.
• Built-infunctionsforcommonstatisticalplots.
• SimplifiedsyntaxcomparedtoMatplotlib.
• EasycustomizationandextensionwithMatplotlib.
ThisbookwillguideyouthroughthemainfunctionalitiesofSeaborn,startingfrom simpleplotsandprogressingtomoreadvancedtopicslikegrids,regressionplots, andcasestudies.
InstallationandRequirements
Beforestarting,youneedtoinstallSeabornanditsdependencies.Werecommend usingavirtualenvironment.
1 pipinstallseabornpandasmatplotlib
Wewillalsouse jupyter toexecuteanddisplayplotsinsidenotebooks:
1 pipinstalljupyter
Recommendedimports:
1 import seabornassns
2 import pandasaspd
3 import matplotlib.pyplotasplt
4
5 sns.set_theme() #Optionalbutrecommended
YourFirstPlot
Let’screateourfirstsimpleplottomakesureeverythingworks:
1 #Createsomedummydata
2 data = pd.DataFrame({
3 "x": range(10),
4 "y":[i **2 for i in range(10)]
5 })
6
7 sns.lineplot(data=data, x="x", y="y") 8 plt.show()
Youshouldseeasimplelineplot.
Figure1: Imagegeneratedbytheprovidedcode. DatasetsWeWillUse
Throughoutthisbook,wewillworkwithseveraldatasets: Penguinsdataset
The penguins datasetisamodernalternativetotheclassicirisdataset.ItprovidesmeasurementsofthreepenguinspeciescollectedfromislandsinPalmer Archipelago,Antarctica.
• Variables:species,island,billlength,billdepth,flipperlength,bodymass, sex,year.
Tipsdataset
Aclassicdatasetaboutrestaurantbillsandtips,includingtotalbill,tip,genderof thecustomer,smokingstatus,dayoftheweek,timeoftheday,andpartysize.
• Variables:total_bill,tip,sex,smoker,day,time,size.
Flightsdataset
Atimeseriesdatasetshowingthenumberofpassengersoveryearsandmonths. Veryusefulforlineplotsandheatmaps.
• Variables:year,month,passengers.
Inthenextchapter,wewillbeginexploring relationalplots,thebackbone ofscatterplots,lineplots,andmore.
RelationalPlots
Introduction
Relationalplotsareusedtodisplayrelationshipsbetweentwoormorevariables. Thisisoneofthemostcommontasksindatavisualization,especiallywhenperformingexploratorydataanalysis(EDA).
Seabornprovidestwomainfunctionsforrelationalplots:- scatterplot() forpoint-basedvisualizations.- lineplot() fortrendortimeseriesvisualizations.
Inthischapter,wewilllearnhowtousethemstepbystep,fromthesimplestusagetoadvancedcustomizationsusingargumentssuchas hue, size,and style. Theseargumentsareespeciallyusefulwhenwewanttoencodeadditionalinformationvisually.
ScatterPlots
Thescatterplotisusedtovisualizetherelationshipbetweentwonumericvariables. Let’sstartwithasimpleexampleusingthe penguins dataset.
1 #Loaddataset
2 penguins = sns.load_dataset("penguins")
3
4 #Basicscatterplot
5 sns.scatterplot(data=penguins,
6 x="bill_length_mm",
7 y="bill_depth_mm")
8 plt.show()
Figure1: Imagegeneratedbytheprovidedcode.
Explanation:
Eachpointrepresentsonepenguin.The x axisshowsthebilllengthandthe y axis thebilldepth.
Addingcolorwith hue
Seabornallowsyoutoeasilymapacategoricalvariabletothecolorofthepoints usingthe hue argument.
1 sns.scatterplot(data=penguins, 2 x="bill_length_mm", 3 y="bill_depth_mm", 4 hue="species") 5 plt.show()
Figure2: Imagegeneratedbytheprovidedcode.
Explanation:
Now,eachspecieshasadi erentcolor.Thismakesiteasiertospotpatternsby species.
Addingsizeandstyle
Youcanaddevenmoreinformationbyusingthe size and style arguments.
1 sns.scatterplot(
2 data=penguins, 3 x="bill_length_mm", 4 y="bill_depth_mm", 5 hue="species",
6 size="flipper_length_mm", 7 style="sex"
8 )
9 plt.show()
Figure3: Imagegeneratedbytheprovidedcode.
Explanation:
• Thesizeofthepointsrepresentstheflipperlength.
• Thestyle(shape)ofthepointsindicatesthesexofeachpenguin.
NoticehowSeabornautomaticallyhandleslegendswhenyouusemultipleencodings.
LinePlots
Lineplotsaremainlyusedtoshowtrendsoveracontinuousvariable,typicallytime orordereddata.
Let’smovetothe flights dataset,whichisperfectforthistypeofvisualization.
1 flights = sns.load_dataset("flights") 2 flights.head()
Basiclineplot
1 sns.lineplot(data=flights, x="year", y="passengers") 2 plt.show()
Figure4: Imagegeneratedbytheprovidedcode.
Explanation:
Thisplotshowstheaveragenumberofpassengersforeachyear,aggregatingall months.
Groupedlineplotwith hue
1 sns.lineplot(data=flights, 2 x="year", 3 y="passengers", IbonMartínez-ArranzPage15
4 hue="month") 5 plt.show()
Figure5: Imagegeneratedbytheprovidedcode.
Explanation:
Nowwearesplittingthedataby month.Eachlinerepresentsamonth,makingit possibletoanalyzeseasonalityandtrendsoveryears.
AdvancedCustomizations
Seabornmakesiteasytocombinemultipleencodings:
Linestylesandmarkers
1 sns.lineplot( 2 data=flights, 3 x="year", 4 y="passengers", 5 hue="month", 6 style="month", 7 markers=True, 8 dashes=False 9 ) 10 plt.show()
Figure6: Imagegeneratedbytheprovidedcode.
Explanation:
• Di erentmonthsnowhavebothdi erentcolorsanddi erentlinestyles (markers).
• Thisisespeciallyhelpfulwhencreatingplotsforprint,wherecoloralone mightnotbesu icient.
ArgumentDescription
hue
Mapsavariabletocolor size
Mapsavariabletomarkersize style
Mapsavariabletomarkerstyle(scatter)or linestyle(lineplot) markers
Addsmarkerstolineplot dashes
Controlswhethertousedashedlines
Whentousescattervs.lineplots?
• Use scatterplot whenyour x variableis notordered (orwhenshowing individualpointsisrelevant).
• Use lineplot whenyour x variableis ordered or continuous,especiallyin timeseries.
• Don’thesitatetocombinebothwhenyouneedtoemphasizebothtrendand individualdatapoints.
Inthenextchapter,wewilldiveinto distributionplots,whichareessential forunderstandingtheshapeofyourdataandidentifyingpatternssuchas skewness,modality,oroutliers.
DistributionPlots
Introduction
Distributionplotsareessentialforunderstandingthestructureofasinglevariable. Theyhelpyouto:
• Checkthedistributionshape(normal,skewed,bimodal).
• Detectoutliers.
• Comparedistributionsbetweengroups.
Seabornprovidesseveralpowerfulfunctionstovisualizedistributions:
• histplot() forhistograms.
• kdeplot() forkerneldensityestimation(KDE).
• ecdfplot() forempiricalcumulativedistributionfunctions.
• rugplot() formarginalticks.
Inthischapter,wewillexploreeachofthesefunctionsandlearnhowtocombine them.
Histogramsareoneofthemostcommontoolstounderstandavariable’sdistribution.
Histogramswith histplot()
BasicHistogram
1 sns.histplot(data=penguins, x="flipper_length_mm") 2 plt.show()
Figure1: Imagegeneratedbytheprovidedcode.
Explanation:
Thehistogramcountshowmanypenguinsfallintoeachbinof flipper_length_mm .Bydefault,Seabornautomaticallychoosesthenumberofbins.
Customizingbins
1 sns.histplot(data=penguins, x="flipper_length_mm", bins =30)
2 plt.show()
Changingthenumberofbinsallowsyoutocontrolthegranularityofthehistogram.
Figure2: Imagegeneratedbytheprovidedcode.
Histogrambycategoryusing hue
1 sns.histplot(data=penguins,
2 x="flipper_length_mm",
3 hue="species",
4 element="step")
5 plt.show()
Figure3: Imagegeneratedbytheprovidedcode.
Explanation:
• The hue argumentallowsyoutoseparatethedistributionbyspecies.
• The element="step" makestheplotlessclutteredbydrawingoutlined
histograms.
KernelDensityEstimationwith kdeplot()
KDEplotsaresmoothedversionsofhistogramsthathelptoseethedistribution shapemoreclearly.
BasicKDE
1 sns.kdeplot(data=penguins, x="flipper_length_mm")
2 plt.show()
Figure4: Imagegeneratedbytheprovidedcode.
Explanation: Thisshowsasmoothedestimateoftheprobabilitydensityfunction.
KDEbygroupwith hue
1 sns.kdeplot(data=penguins, 2 x="flipper_length_mm", 3 hue="species", 4 fill=True) 5 plt.show()
Figure5: Imagegeneratedbytheprovidedcode.
Explanation:
• Byusing hue,youcancomparethedistributionsofflipperlengthsbyspecies.
• fill=True willfilltheareaunderthecurves.
EmpiricalCumulativeDistributionFunctionwith ecdfplot()
AnECDFshowsthecumulativeproportionofthedata.
BasicECDF
1 sns.ecdfplot(data=penguins, x="flipper_length_mm")
2 plt.show()
Figure6: Imagegeneratedbytheprovidedcode.
Explanation:
Foreachvalueof flipper_length_mm,theplotshowstheproportionofpenguinswithflipperlengthslessthanorequaltothatvalue.
ECDFbygroup
1 sns.ecdfplot(data=penguins, 2 x="flipper_length_mm",
3 hue="species")
4 plt.show()
Thisvisualizationishelpfulforcomparingdistributionsacrossgroups.
Figure7: Imagegeneratedbytheprovidedcode.
RugPlotwith rugplot()
Arugplotaddssmalltickmarksalongtheaxistoshowtheactualdatapoints.
BasicRugPlot
1 sns.rugplot(data=penguins, x="flipper_length_mm")
2 plt.show()
Figure8: Imagegeneratedbytheprovidedcode.
Explanation:
Eachsmallverticallinecorrespondstoadatapoint.Itisusefultovisualizethe densityofdatapoints,especiallywhencombinedwithotherplots.
CombiningKDEandRugPlot
1 sns.kdeplot(data=penguins, x="flipper_length_mm")
2 sns.rugplot(data=penguins, x="flipper_length_mm")
3 plt.show()
Figure9: Imagegeneratedbytheprovidedcode.
Explanation:
ThecombinationofKDEandrugplotsallowsyoutoseeboththesmootheddistributionandtheactualdatapoints.
AdditionalTips
• histplot(), kdeplot(),and ecdfplot() acceptcommonarguments like hue, multiple, element, fill,and common_norm.
• Youcaneasilycombinedistributionplotstocreatelayeredvisualizations.
• KDEissensitivetooutliers;useitcarefullywithnoisydata.
SummaryTable PlotPurpose
histplot() Showfrequencycounts(histogram) kdeplot() Estimateandplotthedistribution’sdensity ecdfplot() Displaycumulativedistributionfunction rugplot() Showindividualdatapointsalonganaxis
Inthenextchapter,wewillexplore CategoricalPlots,whichareessential whenworkingwithqualitativedata.
CategoricalPlots
Introduction
Categoricalplotsareamongthemostfrequentlyusedtypesofplotsindataanalysis.Theyareessentialforvisualizingtherelationshipbetweencategoricaland numericalvariablesorforcomparingdistributionsacrossdi erentcategories. Seabornprovidesmultiplefunctionstocreatecategoricalplots,eachwithitsown strengths:
• barplot():Estimateanddisplaythemean(orotherestimator)ofanumericalvariableacrosscategories.
• countplot():Displaycountsofobservationsineachcategoricalbin.
• boxplot():Showdistributionswithquartilesandoutliers.
• violinplot():CombineaboxplotandaKDEforricherdistributioninformation.
• stripplot():Displayallindividualdatapoints,usefulforsmalldatasets.
• swarmplot():Displayindividualpointswithoutoverlap.
Inthischapter,wewillexploretheseplotsindepth.
BarPlotwith barplot()
Barplotsareusedtoshowtheaverageofanumericalvariableforeachcategory, o enwithconfidenceintervals.
Basicbarplot
1 sns.barplot(data=tips, x="day", y="total_bill") 2 plt.show()
Figure1: Imagegeneratedbytheprovidedcode.
Explanation:
Thisplotshowsthe average total_bill foreachdayoftheweek.
Customestimator
Youcanchangetheestimator(bydefaultisthemean)toothersummarystatistics likethemedian:
Figure2: Imagegeneratedbytheprovidedcode.
Barplotwith hue
1 sns.barplot(data=tips, 2 x="day", 3 y="total_bill", 4 hue="sex")
5 plt.show()
Figure3: Imagegeneratedbytheprovidedcode.
Explanation:
Thebarsaresplitby sex withineachday.
CountPlotwith countplot()
Unlike barplot(), countplot() directlycountsthenumberofobservations ineachcategory.
Basiccountplot
1 sns.countplot(data=tips, x="day")
2 plt.show()
Figure4: Imagegeneratedbytheprovidedcode.
Explanation:
Showshowmanyrecordsthereareforeachdayinthedataset.
Countplotwith hue
1 sns.countplot(data=tips, x="day", hue="sex")
2 plt.show()
Figure5: Imagegeneratedbytheprovidedcode.
Thisishelpfulforvisualizinghowthecountsvarybetweensubgroups.
BoxPlotwith boxplot()
Boxplotsshowthedistributionofanumericalvariableusingquartilesandhighlightingpotentialoutliers.
Basicboxplot
1 sns.boxplot(data=tips, x="day", y="total_bill")
2 plt.show()
Figure6: Imagegeneratedbytheprovidedcode.
Explanation:
• Theboxrepresentstheinterquartilerange(IQR).
• Thelineinsidetheboxshowsthemedian.
• The“whiskers”extendto1.5*IQR,andpointsbeyondareconsideredoutliers.
Boxplotwith hue
1 sns.boxplot(data=tips, x="day", y="total_bill", hue="sex" )
2 plt.show()
Figure7: Imagegeneratedbytheprovidedcode.
Adding hue allowsustocomparedistributionsbetweensubgroups.
ViolinPlotwith violinplot()
Violinplotscombinetheinformationofaboxplotandakerneldensityestimation (KDE).
Basicviolinplot
1 sns.violinplot(data=tips, x="day", y="total_bill")
2 plt.show()
Figure8: Imagegeneratedbytheprovidedcode.
Explanation:
Youcanseethedistributionshapeandquartilessimultaneously.
Violinplotwithsplitandhue
1 sns.violinplot(data=tips, 2 x="day", 3 y="total_bill", 4 hue="sex",
6 plt.show()
Figure9: Imagegeneratedbytheprovidedcode.
Thisallowsyoutoseebothdistributionsinthesame“violin”when hue hastwo categories.
Stripplotsarescatterplotswherepointsarearrangedalongacategoricalaxis.
Basicstripplot
1 sns.stripplot(data=tips, x="day", y="total_bill") 2 plt.show()
Figure10: Imagegeneratedbytheprovidedcode.
Explanation:
Usefultoseealldatapoints,especiallywithsmalldatasets.
Stripplotwithjitterandhue
1 sns.stripplot(data=tips, 2 x="day", 3 y="total_bill", 4 jitter=True, 5 hue="sex") 6 plt.show()
• jitter=True spreadsthepointstoavoidoverlap.
• Combiningwith hue helpstodistinguishsubgroups.
Figure11: Imagegeneratedbytheprovidedcode.
SwarmPlotwith swarmplot()
Swarmplotsimproveuponstripplotsbyautomaticallyadjustingpointpositions toavoidoverlaps.
Basicswarmplot
1 sns.swarmplot(data=tips, x="day", y="total_bill")
2 plt.show()
Figure12: Imagegeneratedbytheprovidedcode.
Explanation:
Eachpointrepresentsanobservation,carefullyplacedtoavoidcollisions.
Swarmplotwithhue
1 sns.swarmplot(data=tips, x="day", y="total_bill", hue=" sex")
2 plt.show()
Figure13: Imagegeneratedbytheprovidedcode.
ChoosingtheRightPlot
PlotWhentouse
barplot()
countplot()
boxplot()
violinplot()
stripplot()
swarmplot()
Whenyouwanttoshow aggregatedstatistics (mean, median)bycategory
Whenyouwanttoshow counts ofobservationsper category
Whenyouwanttoshow distributionsummary and detect outliers
Whenyouwanttoshow distributionshape and summarystatistics together
Whenyouwanttoshow alldatapoints forsmall datasets
Sameas stripplot() butbetterhandlingof overlappingpoints
NotesandTips
• Combiningplotsiscommon: boxplot() + stripplot() or violinplot() + swarmplot() createmoreinformativeplots.
• Bemindfulofreadabilitywhenplottingmanycategories.
• Alwayscheckif hue improvesorcluttersyourvisualization.
Inthenextchapter,wewillexplore RegressionPlots,whichareessential whenyouwanttomodelandvisualizelinearornon-linearrelationships betweenvariables.
RegressionPlots
Introduction
Regressionplotsaredesignedtovisualizetherelationshipbetweentwovariables, o entorevealandcommunicatetrendsorpatterns.Theyarecommonlyused to:
• Visualizelinearrelationships.
• Detectnon-linearpatterns.
• Communicatethestrengthofarelationship.
• Showmodel-basedpredictions.
Seabornprovidestwomainfunctionsforregressionplots:
• regplot():Low-levelfunctionthatcreatesascatterplotwitharegression line.
• lmplot():High-levelfunctionwithadditionaloptionsforfacetingandeasy grouping.
SimpleRegressionwith regplot()
Basicregressionplot
1 sns.regplot(data=tips, x="total_bill", y="tip")
2 plt.show()
Figure1: Imagegeneratedbytheprovidedcode.
Explanation:
Thisplotshowsascatterplotof total_bill vs tip withafittedlinearregression lineanda95%confidenceintervalbydefault.
Removingtheconfidenceinterval(ci=None)
1 sns.regplot(data=tips, x="total_bill", y="tip", ci=None)
2 plt.show()
Figure2: Imagegeneratedbytheprovidedcode.
Thissimplifiestheplotbyremovingtheshadedconfidenceinterval.
Changingtheorderoftheregression
Bydefault,Seabornfitsalinearmodel,butyoucanfithigher-orderpolynomials.
1 sns.regplot(data=tips, x="total_bill", y="tip", order=2)
IbonMartínez-ArranzPage53
2 plt.show()
Figure3: Imagegeneratedbytheprovidedcode.
Explanation:
The order=2 argumentfitsaquadraticregressionline.
Robustregression
Youcanmaketheregressionrobusttooutliersbyusing robust=True.
1 sns.regplot(data=tips, x="total_bill", y="tip", robust= True)
2 plt.show()
Figure4: Imagegeneratedbytheprovidedcode.
GroupedRegressionwith lmplot()
Unlike regplot(), lmplot() allowseasygroupingandfaceting.
1 sns.lmplot(data=tips, x="total_bill", y="tip") 2 plt.show()
Figure5: Imagegeneratedbytheprovidedcode.
Producesthesameresultas regplot() butreturnsaFacetGridobject.
Regressionwith hue
1 sns.lmplot(data=tips, x="total_bill", y="tip", hue="sex") 2 plt.show()
Figure6: Imagegeneratedbytheprovidedcode.
Explanation:
Eachgroup(sex)hasitsownregressionlineandscatterpoints.
Changingmarkers
1 sns.lmplot(data=tips, 2 x="total_bill",
3 y="tip", 4 hue="sex", 5 markers=["o", "s"]) 6 plt.show()
Figure7: Imagegeneratedbytheprovidedcode.
Youcanassigndi erentmarkerstoeachgroupforbetterreadability.
Facetingwith col and row
1 sns.lmplot(data=tips,
y="tip",
col="sex",
row="smoker")
plt.show()
Figure8: Imagegeneratedbytheprovidedcode.
Explanation:
Thisisapowerfulfeature:youcansplitthedataintomultiplepanelsbycategories.
MultipleRegression(regressionwithadditional predictors)
Seaborndoesnotdirectlysupportmultipleregressioninthestatisticalsense(with multiple x variables),butyoucanstill:
1.Splitdatausing hue tosimulategroup-wisemultipleregression.
2.Usecolor,size,orstyletoshowadditionalinformation.
Example:Multipleregressionvisualization
1 sns.lmplot(
2 data=tips,
3 x="total_bill",
4 y="tip",
5 hue="sex",
6 col="smoker",
7 scatter_kws={"s":50},
8 line_kws={"linewidth":2}
9 )
10 plt.show()
Figure9: Imagegeneratedbytheprovidedcode.
Explanation:
Thisplotcomparesregressionlinesbyboth sex and smoker status,givinginsight intothecombinede ect.
ArgumentsSummary
ArgumentDescription
order
Polynomialregressionorder ci
hue
Confidenceinterval(default95%)
Groupbyacategoricalvariable robust
Robustregressiontoreduceoutlierinfluence markers
Specifymarkerstyle(s)
ArgumentDescription
col / row
scatter_kws
line_kws
Facetingbycategories(lmplotonly)
Customizescatterplotaesthetics
Customizeregressionlineaesthetics
Whentouse regplot() vs. lmplot()
FunctionRecommendedfor
regplot()
lmplot()
Simpleregressionplotswithoutgroupingor faceting
Regressionplotsinvolvinggrouping(hue)or faceting(col, row)
Notes
• Alwayscheckifalinearmodelisappropriate.Inmanyreal-worldcases, relationshipsmaybenon-linear.
• Robustregressionishelpfulwhenyoususpectoutliersarea ectingthe model.
• Avoidoverfittingwhenincreasing order forpolynomialregressions.
Inthenextchapter,wewillexplore MatrixandHeatmapPlots,which areidealforvisualizingrelationshipsbetweenmanyvariables,especially correlations.
MatrixandHeatmapPlots
Introduction
Matrixplotsareessentialforvisualizing structureddata suchas:
• Correlationmatrices
• Distancematrices
• Contingencytables
• Anytwo-dimensionalarrayofvalues
Seaborno erstwomainfunctionsformatrixvisualizations:
• heatmap():Displaysamatrixwithcoloredcells,optionallyannotated.
• clustermap():Extends heatmap() byapplyingclustering(hierarchical) torowsand/orcolumnsautomatically.
Theseplotsarecommonlyusedtoidentifypatterns,clusters,orstrongrelationships betweenvariables.
TheCorrelationMatrix
Atypicaluseofmatrixplotsistovisualizethe correlation betweennumerical variables.
Let’susethe penguins dataset:
BasicHeatmapwith heatmap()
Figure1: Imagegeneratedbytheprovidedcode.
Explanation:
• Darkercolorsrepresentstrongercorrelations.
• Bydefault,Seabornusesabluecolorpalette.
1 sns.heatmap(corr, annot=True)
2 plt.show()
Figure2: Imagegeneratedbytheprovidedcode.
Explanation:
annot=True displaysthecorrelationcoe icientinsideeachcell.
1 sns.heatmap(corr, annot=True, cmap="coolwarm", center=0) 2 plt.show()
Figure3: Imagegeneratedbytheprovidedcode.
Explanation:
• cmap="coolwarm" helpshighlightpositiveandnegativecorrelations.
• center=0 ensuresthatthecolormapissymmetricaroundzero.
Explanation:
• fmt=".2f" formatsnumberswithtwodecimals.
• linewidths and linecolor addgridlinesbetweencells.
• cbar_kws adjuststhecolorbarsize.
Clustermapwith clustermap()
clustermap() performshierarchicalclusteringofrowsandcolumnstoreveal structuresandgroupsinthedata.
Basicclustermap
1 sns.clustermap(corr) 2 plt.show()
Figure5: Imagegeneratedbytheprovidedcode.
Explanation:
• Variablesarereorderedtoshowclusters.
• Dendrogramsaredisplayedtorepresentthehierarchicalclustering.
Customizingclustermap
1 sns.clustermap(corr, 2 cmap="coolwarm", 3 annot=True, 4 fmt=".2f", 5 linewidths=0.5)
6 plt.show()
Figure6: Imagegeneratedbytheprovidedcode.
Explanation:
Youcanpassmostargumentsfrom heatmap() to clustermap().
Explanation:
• Rowsaremonths,columnsareyears,andthevaluesrepresentthenumber ofpassengers.
• Thisisaperfectexampleofa timematrix.
Clustermaponflightsdata
1 sns.clustermap(flights_pivot, cmap="coolwarm") 2 plt.show()
Figure8: Imagegeneratedbytheprovidedcode.
Explanation:
Thisautomaticallyclustersbothmonthsandyearsbasedonthenumberofpassengers,revealingpatternssuchassimilaryearsorseasonalclusters.
NotesandTips
• heatmap() isgreatfor knownmatrices,likecorrelations.
• clustermap() ishelpfulwhenyoususpectthedatamighthave hidden structure.
• Adjustthe cmap tohighlightthetypeofinformationyouwanttoemphasize (diverging,sequential,etc.).
• Alwayslabelyouraxesclearlywhenreshapingdata.
SummaryTable FunctionPurpose
heatmap()
clustermap()
Plotanynumericmatrixwith customizablecolorsandannotations
Plotamatrixwithhierarchical clusteringappliedautomatically
annot=True Displaynumericalvaluesinsidethe cells
cmap= Controlthecolorpalette fmt= Controlthenumberformattingin annotations
Inthenextchapter,wewilllearnhowtouse FacetGridandPairGrid,two essentialtoolsforbuildingcomplexgridsofplotsautomatically.
MultiplotGrids
Introduction
OneofSeaborn’sstrengthsisitsabilitytoeasilycreate multi-plotgrids,allowing youto:
• Exploredistributionsorrelationshipsacrosssubgroups.
• Createpanelplotswithsharedorindependentaxes.
• Automatecomplexplottinglayouts.
Seabornprovidesthreemaintoolsforthis:
• FacetGrid():Themostflexiblewaytocreatecustomgridsofplots.
• pairplot():Aquickwaytovisualizepairwiserelationshipsbetweenvariables.
• catplot():Combinescategoricalplotswithfacetingcapabilities.
1 g = sns.FacetGrid(tips, col="sex")
2 g.map(sns.histplot, "total_bill")
3 plt.show()
FacetGrid BasicFacetGrid
Figure1: Imagegeneratedbytheprovidedcode.
Explanation:
• Createsonehistogramforeachvalueof sex.
• map() allowsyoutospecifytheplottingfunctiontoapply.
Facetingbyrowandcolumn
1 g = sns.FacetGrid(tips, col="sex", row="smoker")
2 g.map(sns.scatterplot, "total_bill", "tip")
3 plt.show()
Figure2: Imagegeneratedbytheprovidedcode.
Explanation:
• Createsa gridofscatterplots combining sex and smoker.
• Eachcellofthegridcorrespondstoasubgroup.
AddinghueinsideaFacetGrid
1 g = sns.FacetGrid(tips, col="day", hue="sex")
2 g.map(sns.scatterplot, "total_bill", "tip").add_legend()
3 plt.show()
Figure3: Imagegeneratedbytheprovidedcode.
Explanation:
• hue stillworksinsideeachfacet,makingcomparisonseasier.
CustomizingplotsinsideFacetGrid
1 g = sns.FacetGrid(tips, col="day", height=4, aspect=0.7)
2 g.map_dataframe(sns.regplot,
3 x="total_bill", 4 y="tip",
5 scatter_kws={"s":30},
6 line_kws={"color": "red"})
7 g.add_legend()
8 plt.show()
Figure4: Imagegeneratedbytheprovidedcode.
Pairplot
The pairplot() functionisa quickandpowerful waytocreatescatterplot matrices.
BasicPairplot
1 sns.pairplot(penguins.dropna())
2 plt.show()
Figure5: Imagegeneratedbytheprovidedcode.
Explanation:
• Automaticallyplotspairwiserelationshipsbetweenallnumericvariables.
• Diagonalplotsshowthedistributionofeachvariable.
Pairplotwith hue
1 sns.pairplot(penguins.dropna(), hue="species") 2 plt.show()
Figure6: Imagegeneratedbytheprovidedcode.
Explanation:
• Colorstheplotsbyspecies.
IbonMartínez-ArranzPage87
• Veryhelpfulfordetectingclustersandpatterns.
Customizingpairplotmarkers
1 sns.pairplot(penguins.dropna(),
2 hue="species",
3 markers=["o", "s", "D"])
4 plt.show()
Youcanspecifycustommarkersforeachgroup.
Figure7: Imagegeneratedbytheprovidedcode.
Pairplotwithregressionlines
1 sns.pairplot(penguins.dropna(), hue="species", kind="reg" )
2 plt.show()
Figure1: Imagegeneratedbytheprovidedcode.
Explanation:
• Addsregressionlinestoscatterplotsinsteadofplainpoints.
Catplot
catplot() isa general-purpose functionforcategoricalplotswithfacetingbuiltin.
Basiccatplot(equivalenttobarplot)
1 sns.catplot(data=tips, x="day", y="total_bill", kind="bar ")
2 plt.show()
Figure2: Imagegeneratedbytheprovidedcode.
Changingplottype
1 sns.catplot(data=tips, x="day", y="total_bill", kind="box ")
2 plt.show()
Figure3: Imagegeneratedbytheprovidedcode.
Youcanswitchbetween box, violin, strip, swarm,and bar justbychanging the kind.
Facetingwithcatplot
1 sns.catplot(data=tips, 2 x="day", y="total_bill", 3 hue="sex", 4 col="smoker", kind="box") 5 plt.show()
Figure4: Imagegeneratedbytheprovidedcode.
Explanation:
• Createsagridofboxplotssplitby smoker.
• Insideeachplot,the hue separatesthedataby sex.
Changingorientation
1 sns.catplot(data=tips, 2 y="day", x="total_bill", 3 kind="violin", hue="sex", split=True) 4 plt.show()
Figure5: Imagegeneratedbytheprovidedcode.
Byswitching x and y,youchangetheorientationoftheplot.
NotesandTips
• FacetGrid() givesyoufullcontrolbutrequiresmoremanualsetup.
• pairplot() isexcellentforexploratorydataanalysiswithnumericalvariables.
• catplot() isveryflexibleforcategoricaldataandiso entheeasiestchoice forfast,publication-readyplots.
• Use height and aspect toadjustthesizeoffacets.
• Alwayscheckiftoomanyfacetsmaketheplothardertointerpret.
FacetGrid() Fullcontrolforcreatingcustomgrid plots
pairplot() Automaticpairwisescatterplotsfor numericalvariables
catplot() Categoricalplots+automaticfaceting
Inthenextchapter,wewillfocuson AdvancedCustomization,whereyou willlearnhowtochangethemes,palettes,andstylestomakeyourplots publication-quality.
AdvancedCustomization
Introduction
OneofthegreateststrengthsofSeabornisthatitnotonlymakesiteasytogenerate statisticalplotsbutalsoprovidestoolstomakethem visuallyattractiveand publication-ready.
Inthischapter,youwilllearnhowto:-Changeplotthemesandstyles.-Customize colorpalettes.-Adjustcontextstomatchdi erentaudiences(notebooks,papers, presentations).-Combinealltheseoptionsforfullycustomizedvisualizations.
ChangingThemeswith set_theme()
Seaborncomeswithseveralbuilt-inthemesyoucanusetoquicklyadjustthe overallappearanceofyourplots.
Availablethemes:
• "darkgrid" (default)
• "whitegrid"
• "dark"
• "white"
• "ticks"
Example:Switchingthemes
1 sns.set_theme(style="whitegrid")
2 sns.boxplot(data=tips, x="day", y="total_bill")
3 plt.show()
Figure1: Imagegeneratedbytheprovidedcode.
Explanation: whitegrid iscommonlyusedinscientificpublicationsasitprovidescleargridlines.
ManuallyChangingStyleElements
Ifyouwantmorecontrol,youcanadjustspecificstyleelements.
Example:
1 sns.set_style("whitegrid")
2 sns.set_context("talk", font_scale=1.2)
3 sns.boxplot(data=tips, x="day", y="total_bill")
4 plt.show()
Figure2: Imagegeneratedbytheprovidedcode.
• set_context("talk") adaptstheplotforpresentations.
• font_scale allowsyoutoadjustthesizeoftextelementsglobally.
ColorPaletteswith set_palette()
Colorsplayacrucialroleindatavisualization.Seabornprovidespredefinedpalettes andalsoallowsyoutocreatecustomones.
Example:Defaultpalette
1 sns.set_theme()
2 sns.boxplot(data=tips, x="day", y="total_bill", hue="sex" )
3 plt.show()
Figure3: Imagegeneratedbytheprovidedcode.
1 sns.set_palette("pastel")
2 sns.boxplot(data=tips, x="day", y="total_bill", hue="sex" )
3 plt.show()
Figure4: Imagegeneratedbytheprovidedcode.
Example:Usingasequentialpalette
1 sns.set_palette("Blues")
2 sns.violinplot(data=tips, 3 x="day", y="total_bill", 4 hue="sex", split=True) 5 plt.show()
Figure5: Imagegeneratedbytheprovidedcode.
1 custom_palette =["#FF6F61", "#6B5B95"]
2 sns.set_palette(custom_palette)
3 sns.barplot(data=tips, x="day", y="total_bill", hue="sex" )
4 plt.show()
Figure6: Imagegeneratedbytheprovidedcode.
ContextsforDi erentPurposes
Seabornallowsyoutoadjustthe scale ofelementsdependingonyourtarget audience.
Availablecontexts:
• "paper" (smallerplotsforpapers)
• "notebook" (default)
• "talk" (suitableforpresentations)
• "poster" (forlargevisuals)
Example:Switchingcontexts
1 sns.set_context("poster")
2 sns.barplot(data=tips, x="day", y="total_bill", hue="sex" )
3 plt.show()
Figure7: Imagegeneratedbytheprovidedcode.
Explanation:
Noticehowallelements(text,lines,markers)arelarger,suitableforprojecting slidesorposters.
Combiningstyle,palette,andcontext
1 sns.set_theme(style="ticks", 2 palette="Set2", 3 context="talk")
4 sns.violinplot(data=tips, 5 x="day", y="total_bill", hue="sex", 6 split=True)
7 plt.show()
Figure8: Imagegeneratedbytheprovidedcode.
Bycombiningstyle,palette,andcontext,youcancreatecustomizedplotsadapted toyourspecificneeds.
TemporaryCustomizationusing with
Youcanapplytemporarychangestoaspecificplotwithouta ectingothers:
1 withsns.axes_style("whitegrid"):
2 sns.boxplot(data=tips, x="day", y="total_bill")
3 plt.show()
Figure9: Imagegeneratedbytheprovidedcode.
Thisishelpfulwhenyouwanttoapplyadi erentstyleforaspecificplotwithout alteringtheglobaltheme.
SummaryTable FunctionPurpose
set_theme() Settheoverallstyleandpalette
set_style() Changethebackgroundandgrid appearance set_palette() Chooseordefinecolorpalettes set_context() Adjustplotsizeandelementsfor di erentoutputs withsns.axes_style() Temporarilyapplyastylefora specificplot
TipsforChoosingStyle,Palette,andContext
• For papers:use style="whitegrid", context="paper",and sequentialordivergingpalettes.
• For notebooks: style="darkgrid" and context="notebook" work wellforquickexplorations.
• For presentations:increase font_scale anduse context="talk" or "poster".
• Choosepalettesthatare:
– Colorblind-friendly.
– Consistentwiththestoryyouwanttotell.
– Notoverloadedwithtoomanycolors.
Inthenextchapter,youwilllearnhowtoadd AnnotationsandDetails to makeyourplotsmoreinformativeandvisuallyappealing.
AnnotationsandDetails
Introduction
Annotationsandsmalldetailso enmakethedi erencebetweenasimpleplotand agreatplot.Theyallowyouto:
• Highlightimportantfindings.
• Improvereadability.
• Guidetheviewertothekeymessages.
• Makeplotssuitableforpresentationsandpublications.
Inthischapter,youwilllearnhowto:
• Customizeaxislabels,titles,legends.
• Addtextannotationsandmarkers.
• Drawreferencelines.
• Controlticksandgridappearance.
Titles,AxisLabelsandLegends
SeabornintegratessmoothlywithMatplotlib,soyoucaneasilyadjustthebasic componentsofyourplots.
Addingtitleandaxislabels
1 sns.scatterplot(data=tips, x="total_bill", y="tip")
2 plt.title("TipsvsTotalBill")
3 plt.xlabel("TotalBill($)")
4 plt.ylabel("Tip($)")
5 plt.show()
Explanation:
Alwayslabelyouraxesandgiveyourplotacleartitle.
Controllingthelegend
1 sns.scatterplot(data=tips, 2 x="total_bill", y="tip", hue="sex")
3 plt.legend(title="Gender", loc="upperleft")
4 plt.show()
Figure1: Imagegeneratedbytheprovidedcode.
Explanation:
Youcancontrolthelegendpositionandtitleusing plt.legend().
AddingTextAnnotations
Youcanuse plt.text() toinserttextanywhereinyourplot.
Example:
1 sns.scatterplot(data=tips, x="total_bill", y="tip")
2 plt.text(40,8, "Hightiphere!", fontsize=12, color="red ")
3 plt.show()
Figure2: Imagegeneratedbytheprovidedcode.
Explanation:
Useannotationstohighlightspecificpointsorregions.
HighlightingPoints
Youcanalsoplotspecificpointswithdi erentaesthetics.
Example:
1 sns.scatterplot(data=tips, x="total_bill", y="tip")
2 plt.scatter(7.4,5.15, s=200,
3 facecolors='none' ,
4 edgecolors='red' ,
5 linewidths=2)
6 plt.text(7.4,5.65, "Outlier?", color="red")
7 plt.show()
Figure3: Imagegeneratedbytheprovidedcode.
Explanation:
Thistechniqueiso enusedtopointoutoutliersorinterestingdatapoints.
1 sns.scatterplot(data=tips, x="total_bill", y="tip")
2 plt.axhline(5, linestyle="--", color="gray")
3 plt.axvline(30, linestyle="--", color="gray")
4 plt.show()
• axhline() drawsahorizontalline.
• axvline() drawsaverticalline.
Figure4: Imagegeneratedbytheprovidedcode. Diagonalorcustomlineswith plot()
1 sns.scatterplot(data=tips, x="total_bill", y="tip")
2 plt.plot([0,50],[0,10], linestyle="--", color="purple" )
3 plt.show()
Figure5: Imagegeneratedbytheprovidedcode.
AddingTextBoxes
Youcancreatetextwithbackgroundusing bbox.
1 sns.scatterplot(data=tips, x="total_bill", y="tip")
2 plt.text(35,2, "Note:\nLowtipszone",
3 fontsize=10,
1 sns.scatterplot(data=tips, x="total_bill", y="tip")
2 plt.xticks([10,20,30,40,50])
3 plt.yticks([2,4,6,8,10])
4 plt.show()
Figure7: Imagegeneratedbytheprovidedcode.
Adjustinggrid
1 sns.set_style("whitegrid")
2 sns.scatterplot(data=tips, x="total_bill", y="tip")
3 plt.grid(True, linestyle="--", linewidth=0.5)
4 plt.show() Page122IbonMartínez-Arranz
Figure8: Imagegeneratedbytheprovidedcode.
1 sns.scatterplot(data=tips, 2 x="total_bill", y="tip", hue="sex")
3 plt.title("TipsvsTotalBillbyGender")
4 plt.xlabel("TotalBill($)")
5 plt.ylabel("Tip($)")
6 plt.axhline(5, linestyle="--", color="gray")
7 plt.axvline(30, linestyle="--", color="gray")
8 plt.text(40,8, "Hightipzone", fontsize=12, color="red" )
9 plt.legend(title="Gender", loc="upperleft")
10 plt.grid(True, linestyle="--", linewidth=0.5)
11 plt.show()
Figure9: Imagegeneratedbytheprovidedcode. Thisexampleputstogetheralltheconceptscoveredinthischapter.
plt.title(), plt.xlabel(), plt.ylabel()
Setplottitleandaxislabels
plt.legend() Customizelegend
plt.text()
Addtextannotations
plt.scatter() Highlightspecificpoints
plt.axhline(), plt.axvline()
plt.grid()
plt.xticks(), plt.yticks()
Addreferencelines
Controlgridappearance
Customizetickmarks
Inthenextchapter,wewilllearnhowtointegrate SeabornwithMatplotlib, givingyoufullcontroloveryourfigureswhenyouneedadvancedcustomizations.
Seaborn+Matplotlib
Introduction
Throughoutthisbook,wehavefocusedonusingSeaborntocreatebeautifuland informativevisualizations.OneofthegreatadvantagesofSeabornisthatitisbuilt ontopofMatplotlib,whichmeansyouareneverlimitedtoSeaborn’sdefault functionality.Infact,combiningSeabornandMatplotlibisthekeytoproducing fullycustomizedplots.
WhileSeabornprovideshigh-levelplottingfunctionswithsmartdefaultsandeasyto-useinterfaces,Matplotlibgivesyou low-levelaccess to:
• Fine-tunethelayoutanddesignofyourplots.
• Controlelementslikeaxes,legends,annotations,andtickmarks.
• Buildcomplexfigurecompositions(e.g.,custommulti-panellayouts).
Inotherwords, Seabornmakesiteasytocreatebeautifulstatisticalplots,and Matplotliballowsyoutomakethemexactlythewayyouwant.
Inthischapter,youwilllearnhowto:
• AccessandmanipulatetheunderlyingMatplotlibaxesandfiguresfrom Seabornplots.
• Combinemultipleplotsintocustomlayouts.
• Addannotations,arrows,referencelines,andotherfinedetails.
• Controllegendplacementandfiguresizesfordi erentusecases.
Masteringthiscombinationwillgiveyoufullcontroloveryourvisualizations, whetheryouarepreparing:
• Aquickexploratoryplot.
• Areportfigure.
• Aslideforapresentation.
• Apublication-qualityfigure.
Let’snowexplorehowSeabornandMatplotlibworktogetherseamlessly.
AccessingAxesandFigures
Example:
1 ax = sns.scatterplot(data=tips, x="total_bill", y="tip")
2 ax.set_title("ScatterplotwithMatplotlibTitle")
3 ax.set_xlabel("TotalBill($)")
4 ax.set_ylabel("Tip($)")
5 plt.show()
Figure1: Imagegeneratedbytheprovidedcode.
Explanation:
• Seabornreturnsthe AxesSubplot objectdirectly.
• Youcanuse set_title(), set_xlabel(), set_ylabel() asyou wouldinpureMatplotlib.
CustomizingTickLabels
1 ax = sns.boxplot(data=tips, x="day", y="total_bill")
2 ax.set_xticklabels(["Thu", "Fri", "Sat", "Sun"], rotation =45)
3 plt.show()
Figure2: Imagegeneratedbytheprovidedcode.
Explanation:
CustomizeticklabelsjustlikeinMatplotlib.
AdjustingPlotSpacing
Whenusingmultipleplots,itiscommontoadjustspacingmanually.
1 fig, axes = plt.subplots(1,2, figsize=(10,4))
2
3 sns.histplot(data=tips, x="total_bill", ax=axes[0])
4 sns.boxplot(data=tips, x="day", y="total_bill", ax=axes [1])
5
6 plt.tight_layout()
7 plt.show()
Figure3: Imagegeneratedbytheprovidedcode.
Explanation:
• plt.subplots() createsmultipleMatplotlibaxes.
• Passeach ax toSeaborn’s ax= argument.
• tight_layout() improvesspacingautomatically.
MultipleSeabornPlotsinOneFigure
1 fig, axes = plt.subplots(2,2, figsize=(12,8))
3 sns.histplot(data=tips, 4 x="total_bill", 5 ax=axes[0,0])
6 sns.boxplot(data=tips, 7 x="day", y="total_bill", 8 ax=axes[0,1])
9 sns.scatterplot(data=tips, 10 x="total_bill", y="tip", hue="sex", 11 ax=axes[1,0])
12 sns.violinplot(data=tips, 13 x="day", y="total_bill", hue="sex", 14 split=True, ax=axes[1,1])
16 plt.tight_layout() 17 plt.show()
Figure4: Imagegeneratedbytheprovidedcode.
Thisisextremelyusefulfor:
• Creatingcustommulti-plotlayouts.
• Preparingfiguresforpapersandreports.
CombiningSeabornandMatplotlibElements
YoucanaddMatplotlibannotations,arrows,andshapesontopofSeabornplots.
1 ax = sns.scatterplot(data=tips, x="total_bill", y="tip")
2
3 #AddMatplotlibelements
4 ax.axhline(5, linestyle="--", color="red") IbonMartínez-ArranzPage133
5 ax.text(40,8, "Hightipszone", fontsize=12, color="red" )
6 ax.annotate("Outlier?",
xy=(7.4,5.15),
xytext=(8.9,6.65),
arrowprops=dict(facecolor='black' ,
arrowstyle="->"))
plt.show()
Figure5: Imagegeneratedbytheprovidedcode.
Explanation:
• axhline(), text(), annotate() arefromMatplotlibbutworkseamlesslyonSeabornaxes.
Fine-TuningLegends
Movingthelegendoutside
1 ax = sns.scatterplot(data=tips,
2 x="total_bill", y="tip", hue="sex")
3 ax.legend(loc="centerleft", bbox_to_anchor=(1,0.5))
4 plt.tight_layout()
5 plt.show()
Figure6: Imagegeneratedbytheprovidedcode.
Thisiso ennecessarywhenplotsaretightorwhenpreparingfiguresforpapers.
SettingFigureSize
AlthoughSeabornautomaticallyadaptstofiguresize,youmaywantfullcontrol.
Example:
1 plt.figure(figsize=(8,6))
2 sns.boxplot(data=tips, x="day", y="total_bill")
3 plt.show()
Figure7: Imagegeneratedbytheprovidedcode.
Alternatively,use subplots() andcontrolindividualaxessizesifneeded.
SummaryTips
• UseSeabornforplottingandMatplotlibfor fine-tuning.
• plt.subplots() + ax=axes[...] givesyoufullcontroloverlayouts.
• YoucanfreelycombineMatplotlibelementslikearrows,text,annotations withSeabornplots.
• Alwaysuse tight_layout() whenbuildingmulti-plotfigures.
SummaryTable TechniquePurpose
set_title(), set_xlabel() Controlaxislabelsandtitles set_xticklabels() Customizeticklabels plt.subplots() Createcustomlayouts tight_layout() Fixspacingautomatically axhline(), annotate(), text() AddMatplotlibelementsontopof Seabornplots
legend() Controllegendpositionand appearance figsize Setfiguresizemanually
Inthenextchapter,wewillapplyeverythinglearnedsofarin CaseStudies, wherewewillproducefull,high-qualityvisualizationsforrealdatasets.
CaseStudies
Introduction
Inthischapter,wewillputeverythingtogetherbysolvingrealisticproblemsusing Seaborn.Wewill:
• Exploredatasetsvisually.
• Combinemultipleplottingtechniques.
• Customizeplotsforclearcommunication.
• Preparepublication-qualityfigures.
Bytheendofthischapter,youwillhaveacompleteworkflow,fromdataexploration tofullycustomizedvisualization.
CaseStudy1:ExploringthePenguinsDataset
The penguins datasetiso enusedtostudyrelationshipsbetweenmorphological measurementsofpenguinsacrossspecies.
InitialExploration
1 sns.pairplot(penguins.dropna(), hue="species") 2 plt.show()
Figure1: Imagegeneratedbytheprovidedcode.
Observation:
• Wecanidentifyclustersofspecies.
• Strongrelationshipsappearbetween flipper_length_mm, bill_length_mm ,and body_mass_g.
AdvancedScatterplotwithRegression
1 sns.lmplot( 2 data=penguins, 3 x="flipper_length_mm", 4 y="body_mass_g", 5 hue="species", 6 height=6, 7 aspect=1.2, 8 markers=["o", "s", "D"] 9 )
10 plt.title("FlipperLengthvsBodyMassbySpecies")
11 plt.xlabel("FlipperLength(mm)")
12 plt.ylabel("BodyMass(g)")
13 plt.tight_layout()
14 plt.show()
• Regressionlineshighlightspeciesdi erences.
• Markersdi erentiatespeciesvisually.
Figure2: Imagegeneratedbytheprovidedcode.
CaseStudy2:TipsDataset-InsightsforaRestaurant Manager
The tips datasetcouldrepresentrealdatafromarestaurantmanagerinterested inunderstandingcustomertippingbehavior.
1 sns.scatterplot(data=tips, 2 x="total_bill", y="tip", hue="day",
3 style="sex")
4 plt.title("TipAmountvsTotalBill")
5 plt.xlabel("TotalBill($)")
6 plt.ylabel("Tip($)")
7 plt.legend(title="Day/Sex")
8 plt.grid(True, linestyle="--", linewidth=0.5)
9 plt.tight_layout()
10 plt.show()
Observation:
• Higherbillstendtoreceivehighertips.
• Therelationshipvariesslightlyacrossdaysandcustomergender.
Figure3: Imagegeneratedbytheprovidedcode.
AverageTipsbyDayandSmokerStatus
1 sns.catplot(data=tips, x="day", y="tip", hue="smoker", 2 kind="box", height=5, aspect=1.5)
3 plt.title("TipsbyDayandSmokerStatus")
4 plt.tight_layout()
5 plt.show()
Figure4: Imagegeneratedbytheprovidedcode.
• Smokersandnon-smokersshowdi erenttippingpatternsdependingonthe day.
CaseStudy3:TimeTrendsintheFlightsDataset
Thisdatasetshowsthemonthlynumberofpassengersonanairlineoverseveral years.
1 flights_pivot = flights.pivot(index="month", 2 columns="year", 3 values="passengers")
5 sns.heatmap(flights_pivot, annot=True, fmt="d", cmap=" YlGnBu")
6 plt.title("MonthlyPassengerCounts(1949-1960)")
7 plt.xlabel("Year")
8 plt.ylabel("Month")
9 plt.tight_layout()
10 plt.show()
Figure5: Imagegeneratedbytheprovidedcode.
• Clearseasonalityisvisible.
• Passengercountsincreaseyearoveryear.
ClusteredHeatmap
1 sns.clustermap(flights_pivot, cmap="coolwarm")
2 plt.show()
Figure6: Imagegeneratedbytheprovidedcode.
• Groupssimilarmonthsandyearsautomatically.
• Highlightspotentialpatternsandsimilarities.
CaseStudy4:CreatingaPublication-Ready
Let’scombineseveralplotsintoasinglefigureasyouwouldforareportorpaper.
1 fig, axes = plt.subplots(2,2, figsize=(12,10))
sns.scatterplot(data=tips,
x="total_bill", y="tip", hue="day",
ax=axes[0,0]) 7 axes[0,0].set_title("TotalBillvsTip")
x="day", y="tip", hue="smoker",
ax=axes[0,1])
axes[0,1].set_title("TipsbyDayandSmoker")
#Regressionplot
sns.regplot(data=penguins,
x="flipper_length_mm", y="body_mass_g",
ax=axes[1,0]) 19 axes[1,0].set_title("FlipperLengthvsBodyMass")
sns.heatmap(flights_pivot,
annot=True,
fmt="d",
cmap="coolwarm",
linewidths=0.5,
linecolor="white",
annot_kws={"size":8},
ax = axes[1,1])
axes[1,1].set_title("FlightsHeatmap")
32 plt.tight_layout() 33 plt.show()
Figure7: Imagegeneratedbytheprovidedcode.
Notes
• Combiningdi erentSeabornplotsintoa singlefigure isapowerfulwayto communicatemultipleaspectsofyourdata.
• plt.subplots() + ax=axes[...] +Seabornplotsallowfortotalflexibility.
• Donotforgettoadjust titles, legends, fontsizes,and spacing whenpreparingfiguresforreportsorpublications.
KeyTakeaways
• UseSeabornforitssimplebutpowerfulplottingfunctions.
• UseMatplotlibtocustomizedetailsandtoarrangeplotsasneeded.
• Combinebothlibrariestocreatecomplex,yetclearandinformativefigures.
Inthenextchapter,wewillsummarizeuseful TipsandTricks toimprove yourplotsevenmoreandavoidcommonpitfalls.
TipsandTricks
Introduction
Inthischapter,wewillshareacollectionof practicaltips, commonpitfalls,and bestpractices thatwillhelpyou:
• Improvethequalityofyourplots.
• SavetimewhenworkingwithSeaborn.
• Avoidcommonmistakes.
• Prepareplotsfornotebooks,presentations,orpublications.
Theserecommendationsarebasedonreal-worldexperienceandareapplicable regardlessofyourlevelofexpertise.
PlottingTips
Settingaglobalstyleavoidsinconsistentplotsacrossyournotebookorreport.
Tip2—Adjustfiguresizeearlywhenneeded
1 plt.figure(figsize=(8,6))
Thisavoidshavingtoresizeplotslaterorhavingplotstoosmallwhenexporting.
Tip3—Use hue, style,and size meaningfully
Theseargumentsarepowerfulbutcanmakeplotstoocrowdedifoverused. Alwaysaskyourself: Doesaddingcolororstyleimprovereadability?
Tip4—Use tight_layout() o en
1 plt.tight_layout()
Ithelpsautomaticallyfixspacingissues,especiallywhencombiningmultiple plots.
CustomizationTricks
Seabornhandlestheplot,butMatplotlibisyourbestallyforannotations,arrows, textboxes,orpreciselayoutadjustments.
1 ax = sns.scatterplot(data=tips, x="total_bill", y="tip")
2 ax.annotate("Interestingpoint",
3 xy=(40,8),
4 xytext=(30,9),
5 arrowprops=dict(arrowstyle="->"))
Trick2—Controllegendlocationmanually
1 plt.legend(loc="centerleft", bbox_to_anchor=(1,0.5))
Movinglegendsoutsidetheplotiso ennecessaryinpapersandslides.
Trick3—Combinemultipleplotsinonefigure
1 fig, axes = plt.subplots(1,2, figsize=(12,5))
2 sns.boxplot(data=tips, x="day", y="tip", ax=axes[0])
3 sns.violinplot(data=tips, x="day", y="tip", ax=axes[1])
4 plt.tight_layout()
5 plt.show()
Itiso enclearertoshowtwocomplementaryplotssidebysidethantooverload one.
InterpretationTips
• Avoid overplotting:Don’tusetoomanyvariablesencodedvia hue, size, and style simultaneously.
• Alwayscheckiftheplothelpsansweryourquestion.
• Use colorcarefully:
– Avoidunnecessarycolors.
– Usecolorblind-friendlypalettes(e.g., "colorblind", "muted", " Set2").
• Customizelabels:change plt.title(), plt.xlabel(),and plt. ylabel() tomakeplotsself-explanatory.
ExportingPlots
Savingplotsproperly
1 plt.savefig("my_plot.png", 2 dpi=300, 3 bbox_inches="tight")
• Alwaysexportplotswithhighresolution(dpi=300)ifyouplantousethem inpapers.
• Use bbox_inches="tight" toremoveextrawhitespace.
Avoidexportingscreenshots:
Alwaysexportusing savefig() ratherthanscreenshotstopreservequalityand dimensions.
AvoidingCommonMistakes
MistakeRecommendation
UsingdefaultlabelsAlwaysaddinformativetitlesandaxislabels
OverloadingwithvariablesUse hue, style,or size selectively IgnoringlegendsAlwaysadjustorcleanyourlegends
Forgettingtoadjustfigure size Use figsize or set_context() appropriately
Notcheckingcolorblind accessibility
Preferpaletteslike "colorblind" or "muted"
FinalThoughts
Seabornisdesignedtohelpyou:
• Produceinformative,clean,andbeautifulplotswithminimalcode.
• CombineeasilywithMatplotlibforadvancedcontrol.
• Focusonthestoryyouwanttotellthroughyourdata.
Visualizationisnotjustaboutmakingplots;it’saboutmaking insightsvisible.
Inthenext(andfinal)chapter,wewillprovidealistof Referencesand FurtherReading tohelpyoucontinuelearningandimprovingyourdata visualizationskills.
ReferencesandFurtherReading
O icialDocumentation
Thefollowingresourcesareo icialandmaintainedbytheSeabornandMatplotlib developers:
• SeabornO icialDocumentation https://seaborn.pydata.org/
• MatplotlibO icialDocumentation https://matplotlib.org/stable/index.html
• PandasO icialDocumentation https://pandas.pydata.org/docs/
• PythonDataScienceHandbook(JakeVanderPlas) https://jakevdp.github.io/PythonDataScienceHandbook/ RecommendedBooks
• DataVisualizationwithPythonandSeaborn
By:MarcGarcia
Afocusedbookonproducinge ectivevisualizationsusingSeaborn.
• PythonDataScienceHandbook
By:JakeVanderPlas
AcompleteguidetodatasciencewithPython,includingMatplotliband Seaborn.
• StorytellingwithData
By:ColeNussbaumerKnaflic
Highlyrecommendedtoimproveyourabilitytocommunicateinsights throughvisuals.
• FundamentalsofDataVisualization
By:ClausO.Wilke
Open-accessbookexplaininggoodpracticesindatavisualization.
https://clauswilke.com/dataviz/
TutorialsandBlogs
• SeabornTutorials
https://seaborn.pydata.org/tutorial.html
• DataCampSeabornTutorial
https://www.datacamp.com/tutorial/seaborn-python-tutorial
• PracticalBusinessPythonBlog https://pbpython.com/
• TowardsDataScience(VisualizationCategory) https://towardsdatascience.com/tagged/data-visualization
ColorandStyleResources
• ColorBrewer2 (Colorblind-friendlypalettes) https://colorbrewer2.org/
• AdobeColorWheel (Forcreatingcustompalettes) https://color.adobe.com/
• SeabornColorPaletteReference
https://seaborn.pydata.org/tutorial/color_palettes.html
Tomasterdatavisualization:
1.Practicebyreproducingplotsfromarticles,papers,oronlinetutorials.
2. Analyzefiguresinscientificpublicationsandthinkaboutwhatworksand whatcouldbeimproved.
3.Experimentwithdi erentdatasetsandexploretheirrelationshipsvisually.
4.CombineSeabornwithMatplotlibwhenyouneedfullcontrol.
FinalWords
Visualizationisakeyskillforanydatascientist,analyst,orresearcher.Seaborn providesagentlelearningcurvewithpowerfultools,butalwaysremember: “Agoodplotisnotonlyinformativebutalsotellsastory.”
Keepexploring,experimenting,andrefiningyourvisualstorytellingskills! Thankyouforfollowingthisguide!