Data file: Download the and place to Data folder two files: issues_train.tsv and the issues_test.tsv<\/p>\n
GitHubIssueData.cs<\/p>\n
\r\n\/\/ <SnippetAddUsings>\r\nusing Microsoft.ML.Data;\r\n\/\/ <\/SnippetAddUsings>\r\n\r\nnamespace GitHubIssueClassification\r\n{\r\n \/\/ <SnippetDeclareTypes>\r\n public class GitHubIssue\r\n {\r\n [LoadColumn(0)]\r\n public string ID { get; set; }\r\n [LoadColumn(1)]\r\n public string Area { get; set; }\r\n [LoadColumn(2)]\r\n public string Title { get; set; }\r\n [LoadColumn(3)]\r\n public string Description { get; set; }\r\n }\r\n\r\n public class IssuePrediction\r\n {\r\n [ColumnName("PredictedLabel")]\r\n public string Area;\r\n }\r\n \/\/ <\/SnippetDeclareTypes>\r\n}\r\n<\/pre>\nProgram.cs<\/p>\n
\r\n\/\/ <SnippetAddUsings>\r\nusing System;\r\nusing System.IO;\r\nusing System.Linq;\r\nusing Microsoft.ML;\r\n\/\/ <\/SnippetAddUsings>\r\n\r\nnamespace GitHubIssueClassification\r\n{\r\n class Program\r\n {\r\n \/\/ <SnippetDeclareGlobalVariables>\r\n private static string _appPath => Path.GetDirectoryName(Environment.GetCommandLineArgs()[0]);\r\n private static string _trainDataPath => Path.Combine(_appPath, "..", "..", "..", "Data", "issues_train.tsv");\r\n private static string _testDataPath => Path.Combine(_appPath, "..", "..", "..", "Data", "issues_test.tsv");\r\n private static string _modelPath => Path.Combine(_appPath, "..", "..", "..", "Models", "model.zip");\r\n\r\n private static MLContext _mlContext;\r\n private static PredictionEngine<GitHubIssue, IssuePrediction> _predEngine;\r\n private static ITransformer _trainedModel;\r\n static IDataView _trainingDataView;\r\n \/\/ <\/SnippetDeclareGlobalVariables>\r\n static void Main(string[] args)\r\n {\r\n \/\/ Create MLContext to be shared across the model creation workflow objects\r\n \/\/ Set a random seed for repeatable\/deterministic results across multiple trainings.\r\n \/\/ <SnippetCreateMLContext>\r\n _mlContext = new MLContext(seed: 0);\r\n \/\/ <\/SnippetCreateMLContext>\r\n\r\n \/\/ STEP 1: Common data loading configuration\r\n \/\/ CreateTextReader<GitHubIssue>(hasHeader: true) - Creates a TextLoader by inferencing the dataset schema from the GitHubIssue data model type.\r\n \/\/ .Read(_trainDataPath) - Loads the training text file into an IDataView (_trainingDataView) and maps from input columns to IDataView columns.\r\n Console.WriteLine($"=============== Loading Dataset ===============");\r\n\r\n \/\/ <SnippetLoadTrainData>\r\n _trainingDataView = _mlContext.Data.LoadFromTextFile<GitHubIssue>(_trainDataPath, hasHeader: true);\r\n \/\/ <\/SnippetLoadTrainData>\r\n\r\n Console.WriteLine($"=============== Finished Loading Dataset ===============");\r\n\r\n \/\/ <SnippetSplitData>\r\n \/\/ var (trainData, testData) = _mlContext.MulticlassClassification.TrainTestSplit(_trainingDataView, testFraction: 0.1);\r\n \/\/ <\/SnippetSplitData>\r\n\r\n \/\/ <SnippetCallProcessData>\r\n var pipeline = ProcessData();\r\n \/\/ <\/SnippetCallProcessData>\r\n\r\n \/\/ <SnippetCallBuildAndTrainModel>\r\n var trainingPipeline = BuildAndTrainModel(_trainingDataView, pipeline);\r\n \/\/ <\/SnippetCallBuildAndTrainModel>\r\n\r\n \/\/ <SnippetCallEvaluate>\r\n Evaluate(_trainingDataView.Schema);\r\n \/\/ <\/SnippetCallEvaluate>\r\n\r\n \/\/ <SnippetCallPredictIssue>\r\n PredictIssue();\r\n \/\/ <\/SnippetCallPredictIssue>\r\n Console.WriteLine("Press enter to close...");\r\n Console.ReadLine();\r\n }\r\n\r\n public static IEstimator<ITransformer> ProcessData()\r\n {\r\n Console.WriteLine($"=============== Processing Data ===============");\r\n \/\/ STEP 2: Common data process configuration with pipeline data transformations\r\n \/\/ <SnippetMapValueToKey>\r\n var pipeline = _mlContext.Transforms.Conversion.MapValueToKey(inputColumnName: "Area", outputColumnName: "Label")\r\n \/\/ <\/SnippetMapValueToKey>\r\n \/\/ <SnippetFeaturizeText>\r\n .Append(_mlContext.Transforms.Text.FeaturizeText(inputColumnName: "Title", outputColumnName: "TitleFeaturized"))\r\n .Append(_mlContext.Transforms.Text.FeaturizeText(inputColumnName: "Description", outputColumnName: "DescriptionFeaturized"))\r\n \/\/ <\/SnippetFeaturizeText>\r\n \/\/ <SnippetConcatenate>\r\n .Append(_mlContext.Transforms.Concatenate("Features", "TitleFeaturized", "DescriptionFeaturized"))\r\n \/\/ <\/SnippetConcatenate>\r\n \/\/Sample Caching the DataView so estimators iterating over the data multiple times, instead of always reading from file, using the cache might get better performance.\r\n \/\/ <SnippetAppendCache>\r\n .AppendCacheCheckpoint(_mlContext);\r\n \/\/ <\/SnippetAppendCache>\r\n\r\n Console.WriteLine($"=============== Finished Processing Data ===============");\r\n\r\n \/\/ <SnippetReturnPipeline>\r\n return pipeline;\r\n \/\/ <\/SnippetReturnPipeline>\r\n }\r\n\r\n public static IEstimator<ITransformer> BuildAndTrainModel(IDataView trainingDataView, IEstimator<ITransformer> pipeline)\r\n {\r\n \/\/ STEP 3: Create the training algorithm\/trainer\r\n \/\/ Use the multi-class SDCA algorithm to predict the label using features.\r\n \/\/Set the trainer\/algorithm and map label to value (original readable state)\r\n \/\/ <SnippetAddTrainer>\r\n var trainingPipeline = pipeline.Append(_mlContext.MulticlassClassification.Trainers.SdcaMaximumEntropy("Label", "Features"))\r\n .Append(_mlContext.Transforms.Conversion.MapKeyToValue("PredictedLabel"));\r\n \/\/ <\/SnippetAddTrainer>\r\n\r\n \/\/ STEP 4: Train the model fitting to the DataSet\r\n Console.WriteLine($"=============== Training the model ===============");\r\n\r\n \/\/ <SnippetTrainModel>\r\n _trainedModel = trainingPipeline.Fit(trainingDataView);\r\n \/\/ <\/SnippetTrainModel>\r\n Console.WriteLine($"=============== Finished Training the model Ending time: {DateTime.Now.ToString()} ===============");\r\n\r\n \/\/ (OPTIONAL) Try\/test a single prediction with the "just-trained model" (Before saving the model)\r\n Console.WriteLine($"=============== Single Prediction just-trained-model ===============");\r\n\r\n \/\/ Create prediction engine related to the loaded trained model\r\n \/\/ <SnippetCreatePredictionEngine1>\r\n _predEngine = _mlContext.Model.CreatePredictionEngine<GitHubIssue, IssuePrediction>(_trainedModel);\r\n \/\/ <\/SnippetCreatePredictionEngine1>\r\n \/\/ <SnippetCreateTestIssue1>\r\n GitHubIssue issue = new GitHubIssue()\r\n {\r\n Title = "WebSockets communication is slow in my machine",\r\n Description = "The WebSockets communication used under the covers by SignalR looks like is going slow in my development machine.."\r\n };\r\n \/\/ <\/SnippetCreateTestIssue1>\r\n\r\n \/\/ <SnippetPredict>\r\n var prediction = _predEngine.Predict(issue);\r\n \/\/ <\/SnippetPredict>\r\n\r\n \/\/ <SnippetOutputPrediction>\r\n Console.WriteLine($"=============== Single Prediction just-trained-model - Result: {prediction.Area} ===============");\r\n \/\/ <\/SnippetOutputPrediction>\r\n\r\n \/\/ <SnippetReturnModel>\r\n return trainingPipeline;\r\n \/\/ <\/SnippetReturnModel>\r\n }\r\n\r\n public static void Evaluate(DataViewSchema trainingDataViewSchema)\r\n {\r\n \/\/ STEP 5: Evaluate the model in order to get the model's accuracy metrics\r\n Console.WriteLine($"=============== Evaluating to get model's accuracy metrics - Starting time: {DateTime.Now.ToString()} ===============");\r\n\r\n \/\/Load the test dataset into the IDataView\r\n \/\/ <SnippetLoadTestDataset>\r\n var testDataView = _mlContext.Data.LoadFromTextFile<GitHubIssue>(_testDataPath, hasHeader: true);\r\n \/\/ <\/SnippetLoadTestDataset>\r\n\r\n \/\/Evaluate the model on a test dataset and calculate metrics of the model on the test data.\r\n \/\/ <SnippetEvaluate>\r\n var testMetrics = _mlContext.MulticlassClassification.Evaluate(_trainedModel.Transform(testDataView));\r\n \/\/ <\/SnippetEvaluate>\r\n\r\n Console.WriteLine($"=============== Evaluating to get model's accuracy metrics - Ending time: {DateTime.Now.ToString()} ===============");\r\n \/\/ <SnippetDisplayMetrics>\r\n Console.WriteLine($"*************************************************************************************************************");\r\n Console.WriteLine($"* Metrics for Multi-class Classification model - Test Data ");\r\n Console.WriteLine($"*------------------------------------------------------------------------------------------------------------");\r\n Console.WriteLine($"* MicroAccuracy: {testMetrics.MicroAccuracy:0.###}");\r\n Console.WriteLine($"* MacroAccuracy: {testMetrics.MacroAccuracy:0.###}");\r\n Console.WriteLine($"* LogLoss: {testMetrics.LogLoss:#.###}");\r\n Console.WriteLine($"* LogLossReduction: {testMetrics.LogLossReduction:#.###}");\r\n Console.WriteLine($"*************************************************************************************************************");\r\n \/\/ <\/SnippetDisplayMetrics>\r\n\r\n \/\/ Save the new model to .ZIP file\r\n \/\/ <SnippetCallSaveModel>\r\n SaveModelAsFile(_mlContext, trainingDataViewSchema, _trainedModel);\r\n \/\/ <\/SnippetCallSaveModel>\r\n }\r\n\r\n public static void PredictIssue()\r\n {\r\n \/\/ <SnippetLoadModel>\r\n ITransformer loadedModel = _mlContext.Model.Load(_modelPath, out var modelInputSchema);\r\n \/\/ <\/SnippetLoadModel>\r\n\r\n \/\/ <SnippetAddTestIssue>\r\n GitHubIssue singleIssue = new GitHubIssue() { Title = "Entity Framework crashes", Description = "When connecting to the database, EF is crashing" };\r\n \/\/ <\/SnippetAddTestIssue>\r\n\r\n \/\/Predict label for single hard-coded issue\r\n \/\/ <SnippetCreatePredictionEngine>\r\n _predEngine = _mlContext.Model.CreatePredictionEngine<GitHubIssue, IssuePrediction>(loadedModel);\r\n \/\/ <\/SnippetCreatePredictionEngine>\r\n\r\n \/\/ <SnippetPredictIssue>\r\n var prediction = _predEngine.Predict(singleIssue);\r\n \/\/ <\/SnippetPredictIssue>\r\n\r\n \/\/ <SnippetDisplayResults>\r\n Console.WriteLine($"=============== Single Prediction - Result: {prediction.Area} ===============");\r\n \/\/ <\/SnippetDisplayResults>\r\n }\r\n\r\n private static void SaveModelAsFile(MLContext mlContext, DataViewSchema trainingDataViewSchema, ITransformer model)\r\n {\r\n \/\/ <SnippetSaveModel>\r\n mlContext.Model.Save(model, trainingDataViewSchema, _modelPath);\r\n \/\/ <\/SnippetSaveModel>\r\n\r\n Console.WriteLine("The model is saved to {0}", _modelPath);\r\n }\r\n }\r\n}\r\n\r\n<\/pre>\nOutput:<\/p>\n
\r\n* Metrics for Multi-class Classification model - Test Data\r\n*------------------------------------------------------------------------------------------------------------\r\n* MicroAccuracy: 0.739\r\n* MacroAccuracy: 0.674\r\n* LogLoss: .907\r\n* LogLossReduction: .648\r\n<\/pre>\n