/// <summary>
/// Predicts special coverage of the article using built model
/// </summary>
/// <param name="article">Article classified by the method</param>
/// <returns>Special coverage predicted by the model</returns>
public int Classify(Article article)
{
var newX = ProblemBuilder.CreateNode(ArticleFeatures(article), vocabulary.ToList()); //create node
var predictedY = model.Predict(newX); //get special coverage
return((int)predictedY);
}
/// <summary>
/// Constructor that creates object with given training set and testing set
/// </summary>
/// <param name="trainingSet">Training set loaded from a file</param>
/// <param name="testingSet">Testing set loaded from a file</param>
public SVMClassifier(TrainingSet trainingSet, TestingSet testingSet)
{
this.trainingSet = trainingSet;
vocabulary = new HashSet <string>();
x = new List <string>();
y = new List <double>();
foreach (Article article in trainingSet.articles.Values) //load data from the training set
{
string features = ArticleFeatures(article);
//add features and special coverages to lists
AddFeaturesToVocabulary(features);
x.Add(features);
y.Add(article.specialCoverage[0]);
}
foreach (Article article in testingSet.articles.Values) //load articles with given specialCoverage from the testing set
{
if (article.specialCoverage != null)
{
string features = ArticleFeatures(article);
//add features and special coverages to lists
AddFeaturesToVocabulary(features);
x.Add(features);
y.Add(article.specialCoverage[0]);
}
}
//create new problem
ProblemBuilder problemBuilder = new ProblemBuilder();
var problem = problemBuilder.CreateProblem(x, y.ToArray(), vocabulary.ToList());
//create new model using linear kernel
const int C = 1; //C parameter for C_SVC
model = new C_SVC(problem, KernelHelper.LinearKernel(), C);
}
