private protected PredictionEngineBase(IHostEnvironment env, ITransformer transformer, bool ignoreMissingColumns,
SchemaDefinition inputSchemaDefinition = null, SchemaDefinition outputSchemaDefinition = null)
{
Contracts.CheckValue(env, nameof(env));
env.AssertValue(transformer);
Transformer = transformer;
var makeMapper = TransformerChecker(env, transformer);
env.AssertValue(makeMapper);
_inputRow = DataViewConstructionUtils.CreateInputRow <TSrc>(env, inputSchemaDefinition);
PredictionEngineCore(env, _inputRow, makeMapper(_inputRow.Schema), ignoreMissingColumns, outputSchemaDefinition, out _disposer, out _outputRow);
OutputSchema = Transformer.GetOutputSchema(_inputRow.Schema);
}
/// <summary>
/// Sets up the <see cref="IDtoEmotion"/> field.
/// </summary>
/// <param name="inputDataViewSchema">The schema when the initial data view is loaded. This will be inside of <see cref="TrainModel{TFaceData}"/>.</param>
private static void SetupEmotionsToId(DataViewSchema inputDataViewSchema)
{
// Get names of emotions.
var slotNames = new VBuffer <ReadOnlyMemory <char> >();
model.GetOutputSchema(inputDataViewSchema).GetColumnOrNull("Label")?.GetKeyValues(ref slotNames);
IDtoEmotion = new Dictionary <int, string>();
int num = 0;
foreach (var denseValue in slotNames.DenseValues())
{
IDtoEmotion.Add(num++, denseValue.ToString());
}
}
private static void TestSomePredictions(MLContext mlContext)
{
ConsoleHelper.ConsoleWriteHeader("=============== Testing prediction engine ===============");
ITransformer trainedModel = mlContext.Model.Load(ModelPath, out var modelInputSchema);
// Create prediction engine related to the loaded trained model
var predEngine = mlContext.Model.CreatePredictionEngine <InputData, OutputData>(trainedModel);
// Get the key value mapping for Number to Score index
var keyValues = default(VBuffer <float>);
trainedModel.GetOutputSchema(modelInputSchema)["Number"].GetKeyValues <float>(ref keyValues);
var keys = keyValues.Items().ToDictionary(x => (int)x.Value, x => x.Key);
//InputData data1 = SampleMNISTData.MNIST1;
var predictedResult1 = predEngine.Predict(SampleMNISTData.MNIST1);
Console.WriteLine($"Actual: 1 Predicted probability: zero: {predictedResult1.Score[keys[0]]:0.####}");
Console.WriteLine($" One : {predictedResult1.Score[keys[1]]:0.####}");
Console.WriteLine($" two: {predictedResult1.Score[keys[2]]:0.####}");
Console.WriteLine($" three: {predictedResult1.Score[keys[3]]:0.####}");
Console.WriteLine($" four: {predictedResult1.Score[keys[4]]:0.####}");
Console.WriteLine($" five: {predictedResult1.Score[keys[5]]:0.####}");
Console.WriteLine($" six: {predictedResult1.Score[keys[6]]:0.####}");
Console.WriteLine($" seven: {predictedResult1.Score[keys[7]]:0.####}");
Console.WriteLine($" eight: {predictedResult1.Score[keys[8]]:0.####}");
Console.WriteLine($" nine: {predictedResult1.Score[keys[9]]:0.####}");
Console.WriteLine();
var predictedResult2 = predEngine.Predict(SampleMNISTData.MNIST2);
Console.WriteLine($"Actual: 7 Predicted probability: zero: {predictedResult2.Score[keys[0]]:0.####}");
Console.WriteLine($" One : {predictedResult2.Score[keys[1]]:0.####}");
Console.WriteLine($" two: {predictedResult2.Score[keys[2]]:0.####}");
Console.WriteLine($" three: {predictedResult2.Score[keys[3]]:0.####}");
Console.WriteLine($" four: {predictedResult2.Score[keys[4]]:0.####}");
Console.WriteLine($" five: {predictedResult2.Score[keys[5]]:0.####}");
Console.WriteLine($" six: {predictedResult2.Score[keys[6]]:0.####}");
Console.WriteLine($" seven: {predictedResult2.Score[keys[7]]:0.####}");
Console.WriteLine($" eight: {predictedResult2.Score[keys[8]]:0.####}");
Console.WriteLine($" nine: {predictedResult2.Score[keys[9]]:0.####}");
Console.WriteLine();
}
public void RunMultiplePredictions(int numberOfPredictions)
{
var mlContext = new MLContext();
//Load data as input for predictions
IDataView inputDataForPredictions = mlContext.Data.LoadFromTextFile <TransactionObservation>(_dasetFile, separatorChar: ',', hasHeader: true);
Console.WriteLine($"Predictions from saved model:");
ITransformer model = mlContext.Model.Load(_modelfile, out var inputSchema);
var predictionEngine = mlContext.Model.CreatePredictionEngine <TransactionObservation, TransactionFraudPredictionWithContribution>(model);
Console.WriteLine($"\n \n Test {numberOfPredictions} transactions, from the test datasource, that should be predicted as fraud (true):");
mlContext.Data.CreateEnumerable <TransactionObservation>(inputDataForPredictions, reuseRowObject: false)
.Where(x => x.Label == true)
.Take(numberOfPredictions)
.Select(testData => testData)
.ToList()
.ForEach(testData =>
{
Console.WriteLine($"--- Transaction ---");
testData.PrintToConsole();
predictionEngine.Predict(testData).PrintToConsole();
Console.WriteLine($"-------------------");
});
Console.WriteLine($"\n \n Test {numberOfPredictions} transactions, from the test datasource, that should NOT be predicted as fraud (false):");
mlContext.Data.CreateEnumerable <TransactionObservation>(inputDataForPredictions, reuseRowObject: false)
.Where(x => x.Label == false)
.Take(numberOfPredictions)
.ToList()
.ForEach(testData =>
{
Console.WriteLine($"--- Transaction ---");
testData.PrintToConsole();
predictionEngine.Predict(testData).PrintToConsole(model.GetOutputSchema(inputDataForPredictions.Schema));
Console.WriteLine($"-------------------");
});
}
public List <DataStructures.TaxiFarePrediction> RunMultiplePredictions(int numberOfPredictions)
{
// Load data as input for predictions.
IDataView inputDataForPredictions = context.Data.LoadFromTextFile <TaxiTrip>(_datasetFile, hasHeader: true, separatorChar: ',');
Console.WriteLine("Predictions from saved model:");
Console.WriteLine($"\n \n Test {numberOfPredictions} transactions, from the test datasource, that should be predicted as fraud (true):");
var transactionList = new List <DataStructures.TaxiFarePrediction>();
TaxiTripFarePredictionWithContribution prediction;
DataStructures.TaxiFarePrediction explainedPrediction;
context.Data.CreateEnumerable <TaxiTrip>(inputDataForPredictions, reuseRowObject: false)
.Take(numberOfPredictions)
.Select(testData => testData)
.ToList()
.ForEach(testData =>
{
testData.PrintToConsole();
prediction = predictionEngine.Predict(testData);
explainedPrediction = new DataStructures.TaxiFarePrediction(prediction.FareAmount, prediction.GetFeatureContributions(model.GetOutputSchema(inputDataForPredictions.Schema)));
transactionList.Add(explainedPrediction);
});
return(transactionList);
}
