南航暑期国际课程大数据可视化课程大作业总结与试验报告.docx
《南航暑期国际课程大数据可视化课程大作业总结与试验报告.docx》由会员分享，可在线阅读，更多相关《南航暑期国际课程大数据可视化课程大作业总结与试验报告.docx（12页珍藏版）》请在冰豆网上搜索。
南航暑期国际课程大数据可视化课程大作业总结与试验报告
Reportofthecourse
Øtheoreticalsummary
Ihavelearnedsomerelevantknowledgeaboutbigdata,bigdatavisualizationandminingfromthesummercoursewhichisteachedbyProfessorBorisKovalerchuk.Thiscoursedeepensmyunderstandingaboutbigdatavisualizationandminingandbroadensmyeyes.
Inthelecture1,inthefirstplace,itmakesabriefintroductionaboutbigdatavisualizationandmining,includingbigdata,knowledge,dataminingtasks,dataminingtopicsandrelatedfields.Inthefigure1,itdepictstheprocessflowonhowknowledgeisdiscovered.
Figure1knowledgediscoveryprocessflow
Then,ittalksaboutapplicationsofdatamining,whichinvolveinretailindustry,insuranceindustry,telecommunicationsindustryandmanufacturingindustry.Italsodealswiththerulesofdatamining,suchasfirst-orderlogicrules.
Inthesecondplace,inthelecture2,Ilearnedsomeconceptsaboutdatamininganddifferencesbetweendataminingandinformationretrieval.Then,Iknewtherelationshipofdatamininganddatabases.Inthefigure2,itdescribessomeinformationaboutdatabasesanddatamining.
Figure2databasesanddatamining
Afterwards,itintroducesregressionmodelstous,especiallylinearregressionequation,asfollow:
（1）
Besides,multidimensionallinearregressionequationasfollow:
（2）
Atthesametime,Ilearnedtrainingandtestingdataanddecisiontrees.Inthelecture3,itmainlytalksaboutconfusionmatrixandROCcurve.Aconfusionmatrixisavisualizationtooltypicallyusedinsupervisedlearning,whichcontainsinformationaboutactualandpredictedclassificationsdonebyaclassificationsystem.ROCcurveisanotherwaybesidesconfusionmatricestoexaminetheperformanceofclassifiers.Then,itdoeswithTanagradataminingsoftwareandhowtousethesoftware.
Inthelecture4,ittalksabouttheclassificationofdata,whichhasthreeclassificationmethods,namelylineardiscrimination,neuralnetworkanddecisionstree,asfollow:
Figure3lineardiscrimination
Figure4neuralnetwork
Figure5decisionstree
Amongthem,decisionstreeistheimportantpart,whichdealswithchoosingthesplittingattribute,giniindexandinformationgainandgainratio.Theginiindexofthesplitdataisdefinedas
（3）
Gainratioisamodificationoftheinformationgainthatreducesitsbiasonhigh-branchattributes.Itisdefinedas
（4）
Meanwhile,parallelcoordinatetechniquesandvisualizationarediscussedinthelecture.Wecanrepresentdatain1,2,and3-D,evenin4+dimensions.Thesetwopicturesareexamplesofvisualization:
oneisbadvisualization,theotherisbettervisualization.
Figure6badvisualization
Figure7bettervisualization
Inthelecture5,atfirst,itintroducesvisualdataminingsoftwaretous,namelyLotsofLines.Then,ittalksaboutvisualizationofmultidimensionaldatawithcollocatedpairedcoordinatesandgenerallinecoordinates.
Inthethirdplace,inthelecture6,Ilearnedmachinelearning,includingsupervisedmachinelearningandunsupervisedmachinelearning.AndIlearnedthemethodofdata-visualizationandcomparingvisualization.Then,ittalksaboutclusteringandthemethodofclusteringwhichinvolvesinK-meansclusteringandHierarchicalclustering.IttellsusK-meansclusteringindetail.
Inthelecture7,itintroducescollocatedcoordinatestous.Inthelecture8,itmainlytalksaboutcollaborativelosslessvisualization.Inaddition,itdealswithcollaborativeapproachtoenhancevisualizationandadvantagesoflosslessvisualization.Then,itdemonstratesadvantagesofnewlosslessvisualizationbasedontheideaofcollocatingpairsofthecoordinates.ItstatesapplicationsofCPCstarsanddensepixeldisplays.
Finally,inthelecture9,ittalksaboutenvelopesofvisualdatamining.Inthelecture10,theprofessorgaverationalanswersintermsofquestionsthatstudentsputforward.Throughthecourse,Ihaveapreliminaryknowledgeofdatavisualizationandmining.
Øexperimentalreport
1atwo-classconfusionmatrix
（1）total15datarecords：
n
H
W=x
H-100=y
e=x-y
cm
kg
kg
1
171
74
71
3
2
176
62
76
-14
3
181
85
81
4
4
159
50
59
-9
5
167
50
67
-17
6
158
60
58
2
7
160
62
60
2
8
170
56
70
-14
9
178
85
78
7
10
170
55
70
-15
11
177
64
77
-13
12
184
76
84
-8
13
176
82
76
6
14
176
70
76
-6
15
176
71
76
-5
（2）confusionmatrix
Predicted
Negative
Positive
Actual
Negative
7
2
Positive
1
5
（3）computingresults
（1）
FP=
0.222
（2）
FN=
0.167
（3）
TP=
0.833
（4）
TN=
0.778
（5）
AC=
0.800
（6）
P=
0.714
2dataminingbyusingTanagrasoftware
（1）experimentaldata
Outlook
Temp
Humidity
Windy
Class
sunny
75
70
yes
Play
sunny
80
90
yes
DontPlay
sunny
85
85
no
DontPlay
sunny
72
95
no
DontPlay
sunny
69
70
no
Play
overcast
72
90
yes
Play
overcast
83
78
no
Play
overcast
64
65
yes
Play
overcast
81
75
no
Play
rain
71
80
yes
DontPlay
rain
65
70
yes
DontPlay
rain
75
80
no
Play
rain
68
80
no
Play
rain
70
96
no
Play
（2）horizontalaxisisHumidity,verticalaxisisTemp,legendisAttributevalue,attributesareOutlook,WindyandClass.Itsresultsofdataminingarethefollowingfigures.
Figure8itsattributeisOutlook
Figure9itsattributeisWindy
Figure10itsattributeisClass
（3）otherresultsarethefollowingfigures.
Figure11Windy’sdistribution
Figure12Outlook’sdistribution
Figure13regressionscharacteristics
3visualdataminingbyusingLotsofLinessoftware
（1）experimentaldata
Outlook
Temp
Humidity
Windy
Class
1
75
70
1
DontPlay
1
80
90
0
DontPlay
1
85
85
0
DontPlay
3
72
95
1
DontPlay
3
69
70
1
DontPlay
1
72
90
1
Play
1
83
78
0
Play
2
64
65
1
Play
2
81
75
0
Play
2
71
80
1
Play
2
65
70
0
Play
3
75
80
0
Play
3
68
80
0
Play
3
70
96
0
Play
（2）theresultsofdatamining
Figure14graphicsinfourcoordinates
Figure15graphicswithdatatable
升级会员 