static void Main(string[] args)
{
//Load the trainingdata
LoadTrainingData();
//(Instead of shit say poo) Create object of MLContext
var mlContext = new MLContext();
//Step 3 Convert classes with data to IDataView
IDataView trainingDataView = mlContext.Data.LoadFromEnumerable(trainingData);
IDataView testDataView = mlContext.Data.LoadFromEnumerable(testData);
//Create pipeline and define workflow
var pipeline = mlContext.Transforms.Text.FeaturizeText("Features", "FeedbackText")
.Append(mlContext.BinaryClassification.Trainers.FastTree(numberOfLeaves: 50, numberOfTrees: 50, minimumExampleCountPerLeaf: 1));
//Training the algorithm
var model = pipeline.Fit(trainingDataView);
//Load and run TestData
LoadTestData();
var predictions = model.Transform(testDataView);
var metrics = mlContext.BinaryClassification.Evaluate(predictions, "Label");
//Loop added so the code reruns without restarting every time
string strcont = "Y";
while (strcont == "Y")
{
//Reads feedback to string and predicts if IsGood = true or false
Console.WriteLine("Enter feedback: ");
string feedbackString = Console.ReadLine().ToString();
var predictionFunction = mlContext.Model.CreatePredictionEngine <FeedbackTrainingData, FeedbackPrediction>(model);
var feedbackInput = new FeedbackTrainingData();
feedbackInput.FeedbackText = feedbackString;
var feedbackPredicted = predictionFunction.Predict(feedbackInput);
//Prints prediction together with accuracy
string feedbackMessage;
double accuracyToPercent = Convert.ToDouble(metrics.Accuracy) * 100;
if (feedbackPredicted.IsGood == true)
{
feedbackMessage = "Feedback is positive with " + accuracyToPercent + "% accuracy";
}
else
{
feedbackMessage = "Feedback is negative with " + accuracyToPercent + "% accuracy";
}
Console.WriteLine(feedbackMessage);
}
Console.Read();
}
static void Main(string[] args)
{
// Step 1 :- We need to load the training data
//trainingData = SprinklerLab.LoadTrainingData(trainingData);
trainingData.AddRange(ComfortLab.ReadDataForPassXDays());
//System.Console.WriteLine(trainingData.Count);
//SprinklerLab.LoadTrainingDataFromFile(trainingData);
// Step 2 :- Create object of MLContext
var mlContext = new MLContext();
// Step 3 :- Convert your data in to IDataView
IDataView dataView = mlContext.Data.LoadFromEnumerable <FeedbackTrainingData>(trainingData);
var model = ComfortLab.TrainMachine(mlContext, dataView);
var modelHeat = ComfortLab.TrainHeatMachine(mlContext, dataView);
//// Step 4 :- We need to create the pipeline and define the workflows in it.
//var pipeline = mlContext.Transforms.CopyColumns(outputColumnName: "Label", inputColumnName: nameof(FeedbackTrainingData.TurnOnSprinklers))
// .Append(mlContext.Transforms.Concatenate("Features", nameof(FeedbackTrainingData.FeedbackRainfall)))
// .Append(mlContext.BinaryClassification.Trainers.FastTree(numberOfLeaves: 50, numberOfTrees: 50, minimumExampleCountPerLeaf: 1));
//// Step 5 :- Train the algorithm and we want the model out
//var model = pipeline.Fit(dataView);
// Step 6 :- Load the test data and run the test data to check the models accuracy
testData = ComfortLab.LoadTestData(testData);
testData = ComfortLab.LoadTestDataFromFile(testData);
IDataView testDataView = mlContext.Data.LoadFromEnumerable <FeedbackTrainingData>(testData);
var predictions = model.Transform(testDataView);
var heatPredictions = modelHeat.Transform(testDataView);
var metrics = mlContext.BinaryClassification.Evaluate(predictions, "Label");
var heatMetrics = mlContext.BinaryClassification.Evaluate(heatPredictions, "Label");
System.Console.WriteLine($"metrics accuracy: {metrics.Accuracy}");
System.Console.WriteLine($"Accuracy: {metrics.Accuracy:P2}");
System.Console.WriteLine($"Auc: {metrics.AreaUnderRocCurve:P2}");
System.Console.WriteLine($"F1Score: {metrics.F1Score:P2}");
System.Console.WriteLine(Environment.NewLine);
System.Console.WriteLine($"Heat metrics accuracy: {heatMetrics.Accuracy}");
System.Console.WriteLine($"Heat Accuracy: {heatMetrics.Accuracy:P2}");
System.Console.WriteLine($"Heat Auc: {heatMetrics.AreaUnderRocCurve:P2}");
System.Console.WriteLine($"Heat F1Score: {heatMetrics.F1Score:P2}");
// Step 7 :- use the model
string strcont = "Y";
while (strcont == "Y")
{
System.Console.WriteLine("Enter outdoor temp.");
string outdoor = System.Console.ReadLine().ToString();
float output;
float.TryParse(outdoor, out output);
if (output == 0 && outdoor != "0")
{
strcont = outdoor;
}
System.Console.WriteLine("Enter indoor temp.");
string indoorTemperature = System.Console.ReadLine().ToString();
System.Console.WriteLine("Enter time of day");
string timeOfDay = System.Console.ReadLine().ToString();
// Create single instance of sample data from first line of dataset for model input
ComfortModelInput sampleData = new ComfortModelInput()
{
OutdoorTemp = float.Parse(outdoor),
IndoorTemp = float.Parse(indoorTemperature),
IndoorHumidity = 0F,
TimeOfDay = ((DateTimeOffset)DateTime.Parse(timeOfDay)).ToUnixTimeSeconds()
};
// Make a single prediction on the sample data and print results
var predictionResult = HeatConsumeModel.Predict(sampleData);
System.Console.WriteLine(Environment.NewLine + "Using model to make single prediction -- Comparing actual TurnOnHeat with predicted TurnOnHeat from sample data...\n\n");
System.Console.WriteLine($"OutdoorTemp: {sampleData.OutdoorTemp}");
System.Console.WriteLine($"IndoorTemp: {sampleData.IndoorTemp}");
System.Console.WriteLine($"IndoorHumidity: {sampleData.IndoorHumidity}");
System.Console.WriteLine($"TimeOfDay: {sampleData.TimeOfDay}");
bool heatOn = predictionResult.Score > .5 ? true : false;
System.Console.WriteLine($"\n\nPredicted TurnOnHeat: {heatOn}\n\n");
// Make a single prediction on the sample data and print results
var predictionACResult = ACConsumeModel.Predict(sampleData);
System.Console.WriteLine(Environment.NewLine + "Using model to make single prediction -- Comparing actual TurnOnAC with predicted TurnOnAC from sample data...\n\n");
System.Console.WriteLine($"OutdoorTemp: {sampleData.OutdoorTemp}");
System.Console.WriteLine($"IndoorTemp: {sampleData.IndoorTemp}");
System.Console.WriteLine($"IndoorHumidity: {sampleData.IndoorHumidity}");
System.Console.WriteLine($"TimeOfDay: {sampleData.TimeOfDay}");
bool ACOn = predictionACResult.Score > .5 ? true : false;
System.Console.WriteLine($"\n\nPredicted TurnOnAC: {ACOn}\n\n");
var predictionFunction = mlContext.Model.CreatePredictionEngine <FeedbackTrainingData, FeedbackPrediction>(model);
var predictionFunction2 = mlContext.Model.CreatePredictionEngine <FeedbackTrainingData, FeedbackHeatPrediction>(modelHeat);
var feedbackInput = new FeedbackTrainingData();
try
{
feedbackInput.OutdoorTemp = float.Parse(outdoor);
feedbackInput.IndoorTemp = float.Parse(indoorTemperature);
feedbackInput.TimeOfDay = ((DateTimeOffset)DateTime.Parse(timeOfDay)).ToUnixTimeSeconds();
var feedbackPredicted = predictionFunction.Predict(feedbackInput);
System.Console.WriteLine(Environment.NewLine + $"TurnOnAC: {feedbackPredicted.TurnOnAC} | Prediction: {(System.Convert.ToBoolean(feedbackPredicted.TurnOnAC) ? "Positive" : "Negative")} | Probability: {feedbackPredicted.Probability} ");
System.Console.WriteLine($"TurnOnAC :- {feedbackPredicted.TurnOnAC}");
var feedbackPredicted2 = predictionFunction2.Predict(feedbackInput);
System.Console.WriteLine(Environment.NewLine + $"TurnOnHeat: {feedbackPredicted2.TurnOnHeat} | Prediction: {(System.Convert.ToBoolean(feedbackPredicted2.TurnOnHeat) ? "Positive" : "Negative")} | Probability: {feedbackPredicted2.Probability} ");
System.Console.WriteLine($"TurnOnHeat :- {feedbackPredicted2.TurnOnHeat}");
System.Console.WriteLine(Environment.NewLine);
System.Console.WriteLine("=============== End of process, hit <CTRL-C key to finish ===============");
}
catch { }
}
}
