Wolfram Language Paclet Repository
Community-contributed installable additions to the Wolfram Language
Titanic data classifier evaluation
Introduction
In this notebook we show how to evaluate a classifier using ROC functions.
Load the paclet
In[1]:=
Needs["AntonAntonov`ROCFunctions`"]
Data ingestion
In[6]:=
ExampleData[{"MachineLearning","Titanic"}]
Out[6]=
Titanic Survival
In[18]:=
dsTitanic=ResourceFunction["ExampleDataset"][{"MachineLearning","Titanic"}];dsTitanic[1;;-1;;300]
Out[19]=
Classifier
Make classification data:
data=Map[Most[#]Last[#]&,List@@@Normal[dsTitanic[Values]]];
Split data into training and testing parts:
In[71]:=
SeedRandom[111];{dataTraining,dataTesting}=TakeDrop[data,Floor[Length[data]*0.7]];
Make a classifier:
In[73]:=
cf=Classify[dataTraining,Method"NearestNeighbors"]
Out[73]=
ClassifierFunction
 | |
|
Classification results
Get classification results over test data:
In[74]:=
dsRes=Dataset@Map[Join["died"0,"survived"0,cf[#〚1〛,"Probabilities"],"Predicted"cf[#〚1〛],"Actual"#〚2〛]&,dataTesting];dsRes[1;;4]
Out[75]=
ROC functions application
Convert the classification results into ROC associations:
In[76]:=
thRange=Range[0,1,0.025];lsROCs=Table
[{"died","survived"},Normal@dsRes[All,"Actual"],Map[If[#≥th,"survived","died"]&,Normal[dsRes[All,"survived"]]]],{th,thRange};
| ToROCAssociation |
Make ROC plot:
In[77]:=
| ROCPlot |
Out[77]=
""