public Program(string[] args)
{
var context = new MLContext(seed);
var trainer = new Training(context);
var containsPeopleModel = trainer.TrainModel(BinaryAlgorithm,
new DataControl.TrainingOptions()
{
FeatureColumn = "Features",
LabelColumn = "ContainsPeople",
LearningRate = LearningRate,
MaxIterations = MaxIterations
});
var numberOfPeopleModel = trainer.TrainModel(MulticlassAlgorithm,
new DataControl.TrainingOptions()
{
FeatureColumn = "Features",
LabelColumn = "NumberOfPeople",
LearningRate = LearningRate,
MaxIterations = MaxIterations
}
);
var distanceModel = trainer.TrainModel(MulticlassAlgorithm,
new DataControl.TrainingOptions()
{
FeatureColumn = "Features",
LabelColumn = "Distance",
LearningRate = LearningRate,
MaxIterations = MaxIterations
}
);
Console.WriteLine("=============== Evaluating Model accuracy with Test data===============");
trainer.Evaluate(containsPeopleModel, BinaryAlgorithm, "ContainsPeople");
trainer.Evaluate(numberOfPeopleModel, MulticlassAlgorithm, "NumberOfPeople");
trainer.Evaluate(distanceModel, MulticlassAlgorithm, "Distance");
Console.WriteLine("=============== End of model evaluation ===============");
Console.WriteLine("=============== Consuming Model ===============");
Console.WriteLine();
var predictor = new Predictor(context, containsPeopleModel, numberOfPeopleModel, distanceModel);
List <DataControl.Data> data = new List <DataControl.Data>();
foreach (var file in DataControl.FILES)
{
var tmp = DataControl.LoadTrainingData(file);
data.AddRange(tmp.Skip(tmp.Count / 2).Take(tmp.Count / 2));
}
data = RandomPermutation(data);
int incorrectContainsPeople = 0;
int incorrectNumberOfPeople = 0;
int incorrectDistance = 0;
foreach (var d in data)
{
var result = predictor.Predict(d);
if (result.ActualContainsPeople != result.PredictionContainsPeople)
{
incorrectContainsPeople++;
}
if (result.ActualNumberOfPeople != result.PredictionNumberOfPeople)
{
incorrectNumberOfPeople++;
}
if (result.ActualDistance != result.PredictionDistance)
{
incorrectDistance++;
}
}
Console.WriteLine();
Console.WriteLine("Testing Results");
Console.WriteLine("Incorrect Contains People: {0}/{1} {2}%", incorrectContainsPeople, data.Count, (1 - ((double)incorrectContainsPeople / (double)data.Count)) * 100);
Console.WriteLine("Incorrect Number of People: {0}/{1} {2}%", incorrectNumberOfPeople, data.Count, (1 - ((double)incorrectNumberOfPeople / (double)data.Count)) * 100);
Console.WriteLine("Incorrect Distance: {0}/{1} {2}%", incorrectDistance, data.Count, (1 - ((double)incorrectDistance / (double)data.Count)) * 100);
Console.WriteLine();
Console.WriteLine("=========== Finished Consuming Model =============");
}
