///// <summary>
///// if it's leaf, add childNodes
///// </summary>
///// <param name="root"></param>
///// <param name="data">current table; not necessaryly the original table</param>
///// <param name="valToCheck"></param>
///// <returns></returns>
//private static bool CheckIfIsLeaf_sealIfLeaf(DataTable data, TreeNode root, string valToCheck)
//{
// var isLeaf = true;
// var allEndValues = new List<string>();
// // get all leaf values for the attribute in question
// for (var i = 0; i < data.Rows.Count; i++)
// {
// if (data.Rows[i][root.TableIndex].ToString().Equals(valToCheck))
// {
// allEndValues.Add(data.Rows[i][data.Columns.Count - 1].ToString());
// }
// }
// // check whether all elements of the list have the same value
// if (allEndValues.Count > 0 && allEndValues.Any(x => x != allEndValues[0]))
// {
// isLeaf = false;
// }
// // create leaf with value to display and edge to the leaf
// if (isLeaf)
// {
// root.ChildNodes.Add(new TreeNode(true, allEndValues[0], valToCheck));
// }
// return isLeaf;
//}
private static TreeNode __GetRootNodeFroClass_tblAssumeDwelt(DataTable data, string inEdge)
{
var lastIndex = data.Columns.Count - 1;
return(new TreeNode(
data.Columns[data.Columns.Count - 1].ColumnName
, lastIndex
,
new MyAttribute(
data.Columns[lastIndex].ToString()
,
MyAttribute.GetDifferentAttributeNamesOfColumn(data, lastIndex)
),
inEdge
));
}
private static TreeNode GetRootNode_tblAssumeDefiniteAttrsPurged(DataTable data, string inEdge)
{
///if there is only one col: the class .
///
if (data.Columns.Count == 1)
{
return(__GetRootNodeFroClass_tblAssumeDwelt(data, inEdge));
}
var attributes = new List <MyAttribute>();
var highestInformationGainIndex = -1;
var highestInformationGain = double.MinValue;
// Get all names, amount of attributes and attributes for every column
for (var i = 0; i < data.Columns.Count - 1; i++)
{
var differentAttributenames = MyAttribute.GetDifferentAttributeNamesOfColumn(data, i);
attributes.Add(new MyAttribute(data.Columns[i].ToString(), differentAttributenames));
}
// Calculate Entropy (S)
var tableEntropy = _treed.trainSet._EntropyOfClassX.Entropy(data);
for (var i = 0; i < attributes.Count; i++)
{
attributes[i].InformationGain = _treed._TrainSetX.GetGainRatioForCol(data, i, tableEntropy);
if (attributes[i].InformationGain > highestInformationGain)
{
highestInformationGain = attributes[i].InformationGain;
highestInformationGainIndex = i;
}
}
/// todo : gainRatio might be all nils
///
if (highestInformationGain == 0)
{
return(__GetRootNodeFroClass_tblAssumeDwelt(data, inEdge));
}
return(new TreeNode(attributes[highestInformationGainIndex].Name, highestInformationGainIndex, attributes[highestInformationGainIndex], inEdge));
}
