public static DataSet ConstructFromCsv(string filePath, bool hasClassLabel)
{
var set = new DataSet();
var contents = File.ReadAllText(filePath);
var entries = contents.Split(new[] { '\n' }, StringSplitOptions.RemoveEmptyEntries);
var correctAttrCount = 0;
foreach (var entry in entries)
{
var example = new Example();
var fields = entry.Split(new[] { ',' }, StringSplitOptions.RemoveEmptyEntries);
// if we have a class label, the last field is not an attribute
var attrCount = hasClassLabel ? fields.Length - 1 : fields.Length;
// make sure that all entries have the same number of attributes
if (correctAttrCount == 0)
{
correctAttrCount = attrCount;
}
else if (correctAttrCount != attrCount)
{
throw new Exception("Invalid CSV entry, wrong number of attributes.");
}
// add all attributes to example
for (var i = 0; i < attrCount; i++)
{
var value = fields[i].Trim();
example.Attributes[i] = value;
if (set.Attributes.Count == i)
{
set.Attributes.Add(new DataAttribute(i));
}
set.Attributes[i].Values.Add(value);
}
// add class label if we have it
if (hasClassLabel)
{
example.ClassLabel = fields[attrCount].Trim();
}
set.Examples.Add(example);
}
return set;
}
/// <summary>
/// Classifies test data examples using decision tree, and outputs results to stdout.
/// </summary>
public static void Run(TreeNode decisionTree, DataSet testData)
{
try
{
Console.WriteLine();
foreach (var example in testData.Examples)
{
var classLabel = GetClassLabel(decisionTree, example);
var attrs = example.Attributes.OrderBy(x => x.Key).Select(x => x.Value);
Console.WriteLine("{0} ==> {1}", string.Join(",", attrs), classLabel);
}
}
catch // we will only get an exception if the inputs were invalid
{
throw new Exception("Test data cannot be used with decision tree - one of them is invalid.");
}
}
/// <summary>
/// Construct decision tree from training data
/// </summary>
public static TreeNode ConstructDecisionTree(DataSet trainingData)
{
return LearnInternal(trainingData.Examples, trainingData.Attributes, new List<Example>());
}