private TmpNode FindMaxGainNode(List <TmpNode> tmpNodes)
{
if (tmpNodes.Count == 0)
{
throw new Exception("Missing nodes.");
}
TmpNode newNode = new TmpNode {
Gain = -1.0
};
foreach (var node in tmpNodes)
{
if (newNode.Gain < node.Gain)
{
newNode = node;
}
}
return(newNode);
}
private TreeNode GetRootNode(Data data, string parentName, TreeNode parent)
{
// Obliczanie entropii atrybutu decyzyjnego
double decisionEntropy = CalculateEntropy(data.Rows.Select(node => node.Last()).Count(d => d.ToUpper().Equals("YES")), data.Rows.Count);
int decisionAttributeIndex = data.Rows.FirstOrDefault().Count - 1;
List <TmpNode> nodeGains = new List <TmpNode>();
int propertiesCount = data.Rows.FirstOrDefault().Count;
for (int index = 0; index < propertiesCount - 1; index++)
{
List <TmpNode> attribute = GetListOfAttributesEntropies(data, decisionAttributeIndex, index);
nodeGains.Add(new TmpNode
{
Index = index,
Nodes = attribute,
// Obliczanie przyrostu informacji
Gain = CalculateGain(decisionEntropy, attribute, data)
});
}
TmpNode newNode = FindMaxGainNode(nodeGains);
List <string> attributesNames = new List <string>();
foreach (var dataRow in data.Rows)
{
var found = attributesNames.Any(t => t.ToUpper().Equals(dataRow[newNode.Index].ToString().ToUpper()));
if (!found)
{
attributesNames.Add(dataRow[newNode.Index]);
}
}
return(new TreeNode(false, parentName, data.Columns[newNode.Index], newNode.Index, attributesNames, parent));
}
