topic Re: Confusion matrix colour and count based on different variables in Discussions

Confusion matrix colour and count based on different variables

Agustin — Sun, 11 Jun 2023 11:27:51 GMT

Hi,

I'm working on a confusion matrix using the graph builder. I have 2 columns one is Known class (y-axis) an the other Predicted class (x-axis). I want the numbers inside the boxes to show the number of samples in each box. As shown in picture below.

However, instead of the colour gradient map, I would like to have the main diagonal (True Positives) be one colour and the rest of the boxes a different colour. I have a column that is either 1 or 0 depending on whether the sample is a True Positive or not, however when I add this column to the colour option in Graph Builder this happens:

Is there a way I can have the colours as in the second picture, but the labels of the first picture?

Thank you!

Re: Confusion matrix colour and count based on different variables

ih — Wed, 21 Sep 2022 15:04:34 GMT

You could summarize the table first and then draw your graph using row labels instead of the counts:

Names Default To Here( 1 );

dt = Open( "$Sample_data/big class.jmp" );

random reset(2);
dt << New Column("Age 2", Numeric, "Ordinal", Format("Fixed Dec", 5, 0), Formula(:age[Col Shuffle()]));
dt:age << Set Name("Age 1");
dtSum = (dt << Tabulate(
	Add Table(
		Column Table( Statistics( N ) ),
		Row Table( Grouping Columns( :"Age 2"n, :"Age 1"n ) )
	)
)) << Make Into Data Table;

dtSum << New Column("Is Diagonal", Numeric, "Nominal", Format("Best", 12), Formula(:Age 1 == :Age 2));

dtSum << Graph Builder(
	Size( 525, 450 ),
	Show Control Panel( 0 ),
	Variables( X( :Age 1 ), Y( :Age 2 ), Color( :Is Diagonal ) ),
	Elements( Heatmap( X, Y, Legend( 5 ), Label( "Label by Row" ) ) ),
	SendToReport(
		Dispatch(
			{},
			"400",
			ScaleBox,
			{Legend Model(
				5,
				Properties( 0, {Fill Color( -15790017 )}, Item ID( "0", 1 ) ),
				Properties( 1, {Fill Color( 5 )}, Item ID( "1", 1 ) )
			)}
		)
	)
);

Re: Confusion matrix colour and count based on different variables

Agustin — Wed, 21 Sep 2022 15:14:31 GMT

Thank you for your answer but I don't think this gives the right solution, see below:

Left shows the actual values that should be displayed, right is what the script provided results in.

Re: Confusion matrix colour and count based on different variables

ih — Wed, 21 Sep 2022 15:24:47 GMT

Sorry about that, I missed saving that step. Again here I summarized the table first and then drew the graph using row labels instead of the counts, but this time I actually defined row labels:

Names Default To Here( 1 );

dt = Open( "$Sample_data/big class.jmp" );

random reset(2);
dt << New Column("Age 2", Numeric, "Ordinal", Format("Fixed Dec", 5, 0), Formula(:age[Col Shuffle()]));
dt << Run Formulas;
dt:age << Set Name("Age 1");
dtSum = (sum = dt << Tabulate(
	Add Table(
		Column Table( Statistics( N ) ),
		Row Table( Grouping Columns( :"Age 2"n, :"Age 1"n ) )
	)
)) << Make Into Data Table;

dtSum << New Column("Is Diagonal", Numeric, "Nominal", Format("Best", 12), Formula(:Age 1 == :Age 2));

dtSum:N << Label( 1 );

gb = dtSum << Graph Builder(
	Size( 525, 450 ),
	Show Control Panel( 0 ),
	Variables( X( :Age 1 ), Y( :Age 2 ), Color( :Is Diagonal ) ),
	Elements( Heatmap( X, Y, Legend( 5 ), Label( "Label by Row" ) ) ),
	SendToReport(
		Dispatch(
			{},
			"400",
			ScaleBox,
			{Legend Model(
				5,
				Properties( 0, {Fill Color( -15790017 )}, Item ID( "0", 1 ) ),
				Properties( 1, {Fill Color( 5 )}, Item ID( "1", 1 ) )
			)}
		)
	)
);