ISmBackPropagation.RelevanceMeasure

Syntax

RelevanceMeasure: IBinaryModelRelevanceMeasure;

Description

The RelevanceMeasure property returns binary classification quality criteria.

Comments

Criteria of classification quality are calculated, if explanatory series is binary.

Example

To execute the example, add a link to the Stat system assembly.

Sub UserProc;
Var
    NN: SmBackPropagation;
    bin, x1, x2: Array[60] Of Double;
    Ex: ISlSeries;
    res, i, j: Integer;
    s: String;
    CrossValidation: ICrossValidation;
Begin
    NN := New SmBackPropagation.Create;
    // Generate source data
    bin[0] := 0; x1[0] := 0.42; x2[0]:= -47;
    bin[1] := 1; x1[1] := 0.3; x2[1]:=  57;
    bin[2] := 1; x1[2] := 0.32; x2[2]:= 21;
    bin[3] := 1; x1[3] := 1.5; x2[3]:=  17;
    bin[4] := 1; x1[4] := 0.4;  x2[4]:= 92;
    bin[5] := 0; x1[5] := 1.88; x2[5]:= 20;
    bin[6] := 0; x1[6] := 1.38; x2[6]:= -22;
    bin[7] := 1; x1[7] := 1.92; x2[7]:= 160;
    bin[8] := 0; x1[8] := 1.16; x2[8]:= -29;
    bin[9] := 0; x1[9] := 0.78; x2[9]:= 87;
    bin[10] := 0; x1[10] := 0.9; x2[10]:=   136;
    bin[11] := 1; x1[11] := 0.62; x2[11]:=  -19;
    bin[12] := 1; x1[12] := 1.46; x2[12]:=  127;
    bin[13] := 0; x1[13] := 1.1; x2[13]:=   109;
    bin[14] := 1; x1[14] := 0.22; x2[14]:=  3;
    bin[15] := 1; x1[15] := 1.68; x2[15]:=  119;
    bin[16] := 1; x1[16] := 0.9; x2[16]:=   120;
    bin[17] := 0; x1[17] := 0.92; x2[17]:=  43;
    bin[18] := 1; x1[18] := 1.4; x2[18]:=   -49;
    bin[19] := 0; x1[19] := 0.98; x2[19]:=  117;
    bin[20] := 0; x1[20] := 0.44; x2[20]:=  109;
    bin[21] := 0; x1[21] := 1.48; x2[21]:=  144;
    bin[22] := 0; x1[22] := 1.18; x2[22]:=  86;
    bin[23] := 0; x1[23] := 0.18; x2[23]:=  94;
    bin[24] := 0; x1[24] := 1.5; x2[24]:=   96;
    bin[25] := 0; x1[25] := 0.28; x2[25]:=  139;
    bin[26] := 0; x1[26] := 0.58; x2[26]:=  -9;
    bin[27] := 1; x1[27] := 0.92; x2[27]:=  20;
    bin[28] := 1; x1[28] := 0.18; x2[28]:=  -31;
    bin[29] := 1; x1[29] := 1.62; x2[29]:=  106;
    bin[30] := 0; x1[30] := 1.84; x2[30]:=  2;
    bin[31] := 0; x1[31] := 1.6; x2[31]:=   138;
    bin[32] := 1; x1[32] := 0.14; x2[32]:=  53;
    bin[33] := 0; x1[33] := 1.9; x2[33]:=   23;
    bin[34] := 0; x1[34] := 1.88; x2[34]:=  75;
    bin[35] := 1; x1[35] := 0.14; x2[35]:=  30;
    bin[36] := 0; x1[36] := 0.86; x2[36]:=  45;
    bin[37] := 0; x1[37] := 1.4; x2[37]:= -9;
    bin[38] := 1; x1[38] := 1.54; x2[38]:= -21;
    bin[39] := 0; x1[39] := 0.62; x2[39]:= 75;
    bin[40] := 0; x1[40] := 1.72; x2[40]:= 108;
    bin[41] := 1; x1[41] := 0.6; x2[41]:= 35;
    bin[42] := 1; x1[42] := 0.9; x2[42]:= 29;
    bin[43] := 1; x1[43] := 0.9; x2[43]:= -20;
    bin[44] := 1; x1[44] := 1.62; x2[44]:= 78;
    bin[45] := 0; x1[45] := 0; x2[45] := 152;
    bin[46] := 1; x1[46] := 0.42; x2[46] := -2;
    bin[47] := 1; x1[47] := 1.84; x2[47] := 157;
    bin[48] := 0; x1[48] := 1.82; x2[48] := 119;
    bin[49] := 0; x1[49] := 1.26; x2[49] := -37;
    bin[50] := 1; x1[50] := 1.9; x2[50] := 45;
    bin[51] := 1; x1[51] := 0.16; x2[51] := -46;
    bin[52] := 0; x1[52] := 1.28; x2[52] := -35;
    bin[53] := 0; x1[53] := 1.14; x2[53] := -45;
    bin[54] := 0; x1[54] := 0.34; x2[54] := -53;
    bin[55] := 0; x1[55] := 1.3; x2[55] := 104;
    bin[56] := 1; x1[56] := 0.5; x2[56] := 46;
    bin[57] := 0; x1[57] := 0.22; x2[57] := -41;
    bin[58] := 0; x1[58] := 1.46; x2[58] := 31;
    bin[59] := 0; x1[59] := 0.28; x2[59] := 126;

    // Set explained series
    NN.Dependent.Value := bin;
    // Set explanatory series
    Ex := NN.Explanatories;
    Ex.Add.Value := x1;
    Ex.Add.Value := x2;
    // Number of neurons in internal layer
    NN.Neurons := 22;
    // Number of iterations
    NN.Epoch := 400;
    // Minimum value of delta
    NN.DeltasMinimum := 0.001;
    // Set cross validation parameters
    CrossValidation := NN.CrossValidation;
    CrossValidation.SamplingType := CrossValidationSamplingType.Kfold;
    CrossValidation.NumberOfFolds := 4;
    // Run calculation and display results
    res := NN.Execute;
    If res <> 0 Then
        Debug.WriteLine("Failed");
    Else
        // Display summary results of classification
        Debug.WriteLine(" === Summary results of classification  ===");
        Debug.Indent;
        s := "";
        For i := 0 To NN.ClassificationSummary.GetUpperBound(1) Do
            For j := 0 To NN.ClassificationSummary.GetUpperBound(2) Do
                s := s + NN.ClassificationSummary[i, j].ToString + " ";
            End For;
            Debug.WriteLine(s);
            s := "";
        End For;
        Debug.Unindent;
        // Display classification quality criteria
        Debug.WriteLine("Classification quality criteria");
        Debug.Indent;
        Debug.WriteLine("Overall accuracy: " + NN.RelevanceMeasure.Accuracy.ToString);
        Debug.WriteLine("F - estimate: " + NN.RelevanceMeasure.F1.ToString);
        Debug.WriteLine("Number of true-positive values: " + NN.RelevanceMeasure.TruePositive.ToString);
        Debug.WriteLine("Number of true-negative values: " + NN.RelevanceMeasure.TrueNegative.ToString);
        Debug.WriteLine("Number of false-positive values: " + NN.RelevanceMeasure.FalsePositive.ToString);
        Debug.WriteLine("Number of false-negative values: " + NN.RelevanceMeasure.FalseNegative.ToString);
        Debug.Unindent;
    End If;
End Sub UserProc;

After executing the example the specified data is clustered using back-propagation network and the summary classification results and classification quality criteria are displayed in the console window.