/*
* Takes a Datatable with the training data
* translates the data to ints
* trains using the training data
* The last col of the datatable input is the thing to predicted
*/
public void Train(int index)
{
DataTable dataTable = this.theData;
// Debug.Write("DataTable size: ");
// Debug.Write("Rows: " + dataTable.Rows.Count);
// Debug.Write("Cols: " + dataTable.Columns.Count);
ArrayList inputNames = new ArrayList();
foreach (DataColumn column in dataTable.Columns)
{
inputNames.Add(column.ColumnName);
}
this.toPredict = (string)inputNames[index]; // The column to predict
inputNames.RemoveAt(index); // the data input data (predict column removed)
this.inputNamesArr = (string[])inputNames.ToArray(typeof(string));
// Debug.Write("Input arr size: " + inputNamesArr.Length);
// Using Accord.Statistics.Filters to present the data as integers,
// as integers are more efficient
this.codebook = new Codification(dataTable)
{
DefaultMissingValueReplacement = 0
}; // codebook object that can convert strings to ints, null/missing value will be defaulted to 0
DataTable symbols = codebook.Apply(dataTable); // applying our data to the codebook
int[][] inputs = symbols.ToJagged <int>(inputNamesArr); // The conversion to ints
int[] outputs = symbols.ToArray <int>(toPredict); // The conversion to ints
// Debug.Write("Array size: ");
// Debug.Write("inputs: " + inputs.Length);
// Debug.Write("outputs: " + outputs.Length);
// Debug.Write("Test");
var id3 = new ID3Learning() // the id3 algo
{
Attributes = DecisionVariable.FromCodebook(codebook, inputNamesArr) // the trees decision attributes/headers from excel, second argument could be given saying what columns it should be
};
this.tree = id3.Learn(inputs, outputs); // Learn using the inputs and output defined above
// transform the rules of the tree into a string
DecisionSet treeRules = tree.ToRules();
ruleText = treeRules.ToString(codebook, toPredict,
System.Globalization.CultureInfo.InvariantCulture);
Debug.WriteLine(ruleText);
}
private void button_Click(object sender, RoutedEventArgs e)
{
//Decided against the Iris set which is in Accord
//var iris = new Iris();
//double[][] inputs = iris.Instances;
//int[] outputs = iris.ClassLabels;
string[][] data = DataSet.CustomIris.iris_values.Split(new[] { "\r\n" },
StringSplitOptions.RemoveEmptyEntries).Apply(x => x.Split(','));
//features
double[][] inputs = data.GetColumns(0, 1, 2, 3).To <double[][]>();
//labels
string[] labels = data.GetColumn(4);
//Codebook translates any input into usable (integers) for the tree
//var codebook = new Codification(outputs, inputs);
var cb = new Codification("Output", labels);
int[] outputs = cb.Transform("Output", labels);
DecisionVariable[] features =
{
new DecisionVariable("sepal length", DecisionVariableKind.Continuous),
new DecisionVariable("sepal width", DecisionVariableKind.Continuous),
new DecisionVariable("petal length", DecisionVariableKind.Continuous),
new DecisionVariable("petal width", DecisionVariableKind.Continuous),
};
var decisionTree = new DecisionTree(inputs: features, classes: 3);
var c45learner = new C45Learning(decisionTree);
c45learner.Learn(inputs, outputs);
int[] estimated = decisionTree.Decide(inputs);
double error = new ZeroOneLoss(outputs).Loss(decisionTree.Decide(inputs));
//Why rules?
DecisionSet decisionSet = decisionTree.ToRules();
string ruleText = decisionSet.ToString(cb, "Output",
System.Globalization.CultureInfo.InvariantCulture);
//var tree = new DecisionTree(inputs: features, classes: 3);
#region UI
//Set ouput to UI
tb_output.Text = ruleText;
//Calculate the flowers and input to UI -> TODO Bindings
var setosaCount = 0;
var versicolorCount = 0;
var virginicaCount = 0;
for (int i = 0; i < estimated.Length; i++)
{
if (estimated[i] == 0)
{
setosaCount++;
}
if (estimated[i] == 1)
{
versicolorCount++;
}
if (estimated[i] == 2)
{
virginicaCount++;
}
}
tb_setosa.Text = setosaCount.ToString();
tb_versi.Text = versicolorCount.ToString();
tb_virgi.Text = virginicaCount.ToString();
#endregion UI
}
