private static NaiveBayesCalculator TrainNaiveBayes(List <DataSetValue> trainingValues)
{
Dictionary <string, BucketCount> naiveBayesTrainingDataStructure =
NaiveBayesDataTransform.CountSamples(trainingValues);
double probabilityOfOne = 1.0 * trainingValues.Count(t => t.Output) / trainingValues.Count;
var bayesPredictor = new NaiveBayesCalculator(probabilityOfOne, naiveBayesTrainingDataStructure);
return(bayesPredictor);
}
private static void EvaluateNaiveBayesPredictor(ParserResults testData, NaiveBayesCalculator naiveBayesPredictor)
{
Console.WriteLine("Making predictions...");
uint hits = 0, misses = 0;
uint falsePositives = 0, falseNegatives = 0;
foreach (var testExample in testData.Values)
{
var isOnePrediction = naiveBayesPredictor.CalculatePrediction(testExample);
if (isOnePrediction && testExample.Output)
{
hits++;
}
else if (!isOnePrediction && !testExample.Output)
{
hits++;
}
else if (isOnePrediction && !testExample.Output)
{
misses++;
falsePositives++;
}
else if (!isOnePrediction && testExample.Output)
{
misses++;
falseNegatives++;
}
else
{
throw new InvalidOperationException();
}
}
Console.WriteLine("Score: {0}%. Hits: {1}, Misses: {2}", 100.0 * hits / (misses + hits), hits, misses);
Console.WriteLine("FalsePositives: {0}. FalseNegatives: {1}", falsePositives, falseNegatives);
}
static void Main(string[] args)
{
string errorMessage = "";
if (!File.Exists(TrainingDataPath))
{
errorMessage += $"Failed to find file ${TrainingDataPath} - please update variable ${nameof(TrainingDataPath)} or create that file.\n";
}
if (!File.Exists(TestDataPath))
{
errorMessage += $"Failed to find file ${TestDataPath} - please update variable ${nameof(TestDataPath)} or create that file.\n";
}
if (errorMessage != "")
{
Console.ForegroundColor = ConsoleColor.Red;
Console.WriteLine("Not all files available - not running!");
Console.WriteLine(errorMessage);
Console.ResetColor();
Console.WriteLine("Press any key to continue...");
Console.ReadKey();
return;
}
var startTime = DateTime.Now;
Console.WriteLine(startTime);
Console.WriteLine("Parsing...");
List <EmailExample> trainingEmails;
Dictionary <string, WordCount> trainingCounts = CsvParserUtils.ParseEmailExamples(TrainingDataPath, out trainingEmails);
List <EmailExample> testEmails;
CsvParserUtils.ParseEmailExamples(TestDataPath, out testEmails);
double probabilitySpam = 1.0 * trainingEmails.Count(t => t.IsSpam) / trainingEmails.Count;
Console.WriteLine("Making predictions...");
uint hits = 0, misses = 0;
uint falsePositives = 0, falseNegatives = 0;
foreach (var emailExample in testEmails)
{
//double probabilityOfSpam = NaiveBayesCalculator.ObtainProbabilityOfSpam(emailExample.WordsInEmail, trainingCounts, probabilitySpam);
//bool isSpamPrediction = probabilityOfSpam > 0.5;
var probabilityOfSpam = NaiveBayesCalculator.ObtainProbabilityOfSpam(emailExample.WordsInEmail, trainingCounts, probabilitySpam, trainingCounts.Count);
bool isSpamPrediction = probabilityOfSpam.Item1 > probabilityOfSpam.Item2;
if (isSpamPrediction && emailExample.IsSpam)
{
hits++;
}
else if (!isSpamPrediction && !emailExample.IsSpam)
{
hits++;
}
else if (isSpamPrediction && !emailExample.IsSpam)
{
misses++;
falsePositives++;
}
else if (!isSpamPrediction && emailExample.IsSpam)
{
misses++;
falseNegatives++;
}
else
{
throw new InvalidOperationException();
}
}
Console.WriteLine("Score: {0}%. Hits: {1}, Misses: {2}", 100.0 * hits / (misses + hits), hits, misses);
Console.WriteLine("FalsePositives: {0}. FalseNegatives: {1}", falsePositives, falseNegatives);
var endTime = DateTime.Now;
Console.WriteLine(endTime);
var totalMinutes = (endTime - startTime).TotalMinutes;
Console.WriteLine("Took {0} minutes.", totalMinutes);
Console.WriteLine("Press any key to quit...");
Console.ReadKey();
}
