protected override ITransformer TrainModel(IDataView dataView)
{
//数字键类型列
IEstimator <ITransformer> estimator =
Context.Transforms.Conversion.MapValueToKey(LabelField, nameof(GitHubIssue.Area));
var fields = new[] { nameof(GitHubIssue.Title), nameof(GitHubIssue.Description) };
var features = new List <string>();
//特征列
foreach (var field in fields)
{
var featureField = $"{field}Featurized";
estimator = estimator.Append(Context.Transforms.Text.FeaturizeText(featureField, field));
features.Add(featureField);
}
//合并特征列
estimator = estimator.Append(Context.Transforms.Concatenate(FeaturesField, features.ToArray()));
//缓存数据视图
estimator = estimator.AppendCacheCheckpoint(Context);
//添加训练算法
estimator = estimator
.Append(Context.MulticlassClassification.Trainers.SdcaMaximumEntropy())
.Append(Context.Transforms.Conversion.MapKeyToValue(PredictedLabelField));
Console.WriteLine("=============== Create and Train the Model ===============");
var model = estimator.Fit(dataView);
Console.WriteLine("=============== End of training ===============");
return(model);
}
private IEstimator <ITransformer> SetUpLearningModel()
{
// transform your data
IEstimator <ITransformer> pipeLine = dotNetMachineLearningContext.Transforms.Conversion.MapValueToKey("Label");
pipeLine = pipeLine.Append(dotNetMachineLearningContext.Transforms.Concatenate("Features", "SepalLength", "SepalWidth", "PetalLength", "PetalWidth"));
pipeLine = pipeLine.AppendCacheCheckpoint(dotNetMachineLearningContext);
//// add a learner
//// Add a learning algorithm to the pipeline. e.g.(What type of iris is this?)
//// Assign numeric values to text in the "Label" column,
//// because only numbers can be processed during model training.
pipeLine = pipeLine.Append(dotNetMachineLearningContext.MulticlassClassification.Trainers.SdcaMaximumEntropy(labelColumnName: "Label", featureColumnName: "Features"));
//// Convert the Label back into original text (after converting to number in step 3)
pipeLine = pipeLine.Append(dotNetMachineLearningContext.Transforms.Conversion.MapKeyToValue("PredictedLabel"));
return(pipeLine);
}
private IEstimator <ITransformer> SetUpLearningModel()
{
// transform your data
IEstimator <ITransformer> pipeLine = dotNetMachineLearningContext.Transforms.Conversion.MapValueToKey("WeekDay");
pipeLine = pipeLine.Append(dotNetMachineLearningContext.Transforms.Concatenate("Features", "Temp", "RH"));
// add a learner
pipeLine = pipeLine.AppendCacheCheckpoint(dotNetMachineLearningContext);
// Add a learning algorithm to the pipeline.
// Assign numeric values to text in the labeled column,
// because only numbers can be processed during model training.
pipeLine = pipeLine.Append(dotNetMachineLearningContext.MulticlassClassification.Trainers.SdcaMaximumEntropy(labelColumnName: "WeekDay", featureColumnName: "Features"));
// Convert the Label back into original text (after converting to number in step 3)
pipeLine = pipeLine.Append(dotNetMachineLearningContext.Transforms.Conversion.MapKeyToValue(
outputColumnName: "PredictedLabel",
inputColumnName: "WeekDay"));
return(pipeLine);
}
