private static Task ClassifyAsync(
IModelDefinitionBuilder <SpamInput, MulticlassClassificationFoldsAverageMetricsResult> modelBuilder,
string text,
string expectedResult,
ILogger logger,
CancellationToken cancellationToken)
{
return(Task.Run(
() =>
{
var predictor = modelBuilder.MLContext.Model.CreatePredictionEngine <SpamInput, SpamPrediction>(modelBuilder.Model);
var input = new SpamInput {
Message = text
};
var prediction = predictor.Predict(input);
var result = prediction.IsSpam == "spam" ? "spam" : "not spam";
if (prediction.IsSpam == expectedResult)
{
logger.LogInformation("[ClassifyAsync][Predict] result: '{0}' is {1}", input.Message, result);
}
else
{
logger.LogWarning("[ClassifyAsync][Predict] result: '{0}' is {1}", input.Message, result);
}
},
cancellationToken));
}
public bool isSpam(string text)
{
// Create MLContext
MLContext mlContext = new MLContext();
//Define DataViewSchema for data preparation pipeline and trained model
DataViewSchema modelSchema;
string filePath = Path.Combine(AppDomain.CurrentDomain.BaseDirectory, "Content", "model.zip");
// Load trained model
var model = mlContext.Model.Load(filePath, out modelSchema);
var predictor = mlContext.Model.CreatePredictionEngine <SpamInput, SpamPrediction>(model);
var input = new SpamInput {
Message = text
};
var prediction = predictor.Predict(input);
if (prediction.isSpam == "spam")
{
return(true);
}
return(false);
}
public IActionResult Predict(SpamInput input)
{
var model = new SpamDetectionMLModel();
model.Build();
model.Train();
ViewBag.Prediction = model.Predict(input);
return(View());
}
public static void ClassifyMessage(PredictionEngine <SpamInput, SpamPrediction> predictor, string message)
{
var input = new SpamInput {
Message = message
};
var prediction = predictor.Predict(input);
Console.WriteLine("The message '{0}' is {1}", input.Message, prediction.isSpam ? "spam" : "not spam");
}
private void ClassifySpamMessage(PredictionEngine <SpamInput, SpamPrediction> predictor, string message)
{
var input = new SpamInput {
Message = message
};
var prediction = predictor.Predict(input);
_logger.LogInformation("The message '{0}' is {1}", input.Message, prediction.IsSpam == "spam" ? "spam" : "not spam");
}
public static void ClassifyMessage(PredictionEngine <SpamInput, SpamPrediction> predictor, string message)
{
var input = new SpamInput {
Message = message
};
var prediction = predictor.Predict(input);
var isSpamStr = prediction.IsSpam == "spam" ? "spam" : "not spam";
Log.Information($"The message '{input.ToJson(true)}' is '{isSpamStr}'");
}
public bool IsTrue(SpamInput input)
{
var model = new SpamDetectionMLModel();
model.Build();
model.Train();
var result = model.Predict(input);
return(Convert.ToBoolean(result.isSpam));
}
public static bool PredictMessage(string message)
{
if (SpamEngine == null)
{
Log.Information("SpamEngine need be built.");
SetupEngine();
}
var input = new SpamInput {
Message = message
};
var predict = SpamEngine.Predict(input);
var isSpam = predict.IsSpam == "spam";
Log.Information($"IsSpam: {isSpam}");
return(isSpam);
}
public IActionResult Predict(SpamInput input)
{
var predictVm = new PredictViewModel();
var result = predictVm.IsTrue(input);
if (!ModelState.IsValid)
{
return(View());
}
if (result)
{
ModelState.AddModelError("Message", "Invalid text found");
input.Message = "";
return(View(input));
}
return(View());
}
static void userMessage(string message, PredictionEngine <SpamInput, SpamPrediction> predictionEngine)
{
var messages = new SpamInput[] {
new SpamInput()
{
Message = message
}
};
var myPredictions = from m in messages
select(Message : m.Message, Prediction : predictionEngine.Predict(m));
foreach (var p in myPredictions)
{
Console.WriteLine($" [{p.Prediction.Probability:P2}] {p.Message}");
}
Console.WriteLine("Skriv noget spam eller ham: ");
var userInput = Console.ReadLine();
userMessage(userInput, predictionEngine);
}
static void Main(string[] args)
{
var context = new MLContext();
// data fra spam.csv(træningsæt) bliver læst in i memorien
var data = context.Data.LoadFromTextFile <SpamInput>(
path: dataPath,
hasHeader: true,
separatorChar: ',');
// use 80% for training and 20% for testing
var partitions = context.Data.TrainTestSplit(
data,
testFraction: 0.2);
// Først så laver den spam, eller ham om en til en bool'
// Som vores model kan arbejde med.
// featurized text hjælper med hvordan model skal forstå data
// til sidst bruger jeg en metode, der hjælper på at lave mere præsise forudsigelser
var pipeline = context.Transforms.CustomMapping <FromLabel, ToLabel>(
mapAction: (input, output) => { output.Label = input.RawLabel == "spam" ? true : false; },
contractName: "MyLambda")
.Append(context.Transforms.Text.FeaturizeText(
outputColumnName: "Features",
inputColumnName: nameof(SpamInput.Message)))
.Append(context.BinaryClassification.Trainers.SdcaLogisticRegression());
// her træner jeg min model.
var model = pipeline.Fit(partitions.TrainSet);
// Evaluerer min model
Console.WriteLine("Evalulerer modellen");
var predictions = model.Transform(partitions.TestSet);
var metrics = context.BinaryClassification.Evaluate(
data: predictions,
labelColumnName: "Label",
scoreColumnName: "Score");
Console.WriteLine($" Accuracy: {metrics.Accuracy:P2}");
Console.WriteLine($" Auc: {metrics.AreaUnderRocCurve:P2}");
Console.WriteLine($" Auprc: {metrics.AreaUnderPrecisionRecallCurve:P2}");
Console.WriteLine($" F1Score: {metrics.F1Score:P2}");
Console.WriteLine($" LogLoss: {metrics.LogLoss:0.##}");
Console.WriteLine($" LogLossReduction: {metrics.LogLossReduction:0.##}");
Console.WriteLine($" PositivePrecision: {metrics.PositivePrecision:0.##}");
Console.WriteLine($" PositiveRecall: {metrics.PositiveRecall:0.##}");
Console.WriteLine($" NegativePrecision: {metrics.NegativePrecision:0.##}");
Console.WriteLine($" NegativeRecall: {metrics.NegativeRecall:0.##}");
Console.WriteLine();
Console.WriteLine("Predicter udfald af beskeder");
var predictionEngine = context.Model.CreatePredictionEngine <SpamInput, SpamPrediction>(model);
var messages = new SpamInput[] {
new SpamInput()
{
Message = "Hello, win free iPad"
},
new SpamInput()
{
Message = "Im home in few"
},
new SpamInput()
{
Message = "Hundreds of medications all 80% off or more!"
},
new SpamInput()
{
Message = "CONGRATS U WON LOTERY CLAIM UR 1 MILIONN DOLARS PRIZE"
},
};
var myPredictions = from m in messages
select(Message : m.Message, Prediction : predictionEngine.Predict(m));
foreach (var p in myPredictions)
{
Console.WriteLine($" [{p.Prediction.Probability:P2}] {p.Message}");
}
Console.WriteLine("Skriv noget spam eller ham :");
var userInput = Console.ReadLine();
userMessage(userInput, predictionEngine);
}
public SpamPrediction Predict(SpamInput input)
{
var predictor = mlContext.Model.CreatePredictionEngine <SpamInput, SpamPrediction>(_model);
return(predictor.Predict(input));
}