private double Predict(Message testMessage, MessageClass msgClass, double penalty, double windowWidth)
{
return(Math.Log(penalty * msgClass.Prevalence)
+ _headerDict.Sum(word =>
ParsenRosenblatt(word, testMessage.GetHeaderPrevalence(word), msgClass, windowWidth, msg => msg.GetHeaderPrevalence))
+ _bodyDict.Sum(word =>
ParsenRosenblatt(word, testMessage.GetBodyPrevalence(word), msgClass, windowWidth, msg => msg.GetBodyPrevalence)));
}
public DataSet(IList <string> headerDict, IList <string> bodyDict, MessageClass trainLegit,
MessageClass trainSpam, IList <Message> testLegit, IList <Message> testSpam)
{
_headerDict = headerDict;
_bodyDict = bodyDict;
_trainLegit = trainLegit;
_trainSpam = trainSpam;
_testLegit = testLegit;
_testSpam = testSpam;
}
private static double ParsenRosenblatt(string word, double prevalence, MessageClass msgClass, double windowWidth, Func <Message, Func <string, double> > prev)
{
var likeliness = msgClass.Sum(msg => KernelFunc(Math.Abs(prevalence - prev(msg).Invoke(word)) / windowWidth));
return(Math.Log(likeliness / msgClass.Size));
}
