如何使用ML.net

Question

我有这个：

    class SuggestedFeedBackPredition
    {

        [LoadColumn(0), ColumnName("Label")]
        public bool IsGood { get; set; }
    }

    public class SuggestedFeedbackTrainingData
    {
        [LoadColumn(1), ColumnName("SuggestedFeedBackText")]
        public string? SuggestedFeedBackText { get; set; }

        [LoadColumn(0), ColumnName("Label")]
        public bool IsGood { get; set; }

        public Single Score { get; set; }
        public Single Probability { get; set; }
        public Single PredictedLabel { get; set; }
    }

我有这个：

var mlContext = new MLContext();
        var mlData = mlContext.Data.LoadFromEnumerable<SuggestedFeedbackTrainingData>(suggestedFeedbackTrainingData());
        //var pipline = mlContext.Transforms.Text.FeaturizeText("Label", "Features");
        //var pipline = mlContext.Transforms.Text.FeaturizeText("Features", "SuggestedFeedbBackPridiction");
        //var pipline = mlContext.Transforms.CopyColumns(outputColumnName: "Features", inputColumnName: "SuggestedFeedBackText").Append(mlContext.Transforms.Concatenate("Label",
        //        "Label"));
        var dataProcessPipeline = mlContext.Transforms.Text.FeaturizeText("SuggestedFeedBackText");
        ITransformer model = dataProcessPipeline.Fit(mlData);
        //var dataview1 = mlContext.Data.LoadFromEnumerable<SuggestedFeedbackTrainingData>(testFeedbackTrainingData());
        //var predictions = model.Transform(dataview1);
        //var metrics = mlContext.BinaryClassification.Evaluate(predictions, "IsGood", "Score");

        var predictionsObj = mlContext.Model.CreatePredictionEngine<SuggestedFeedbackTrainingData, SuggestedFeedBackPredition>(model);
        var ff = new SuggestedFeedbackTrainingData();
        ff.SuggestedFeedBackText = text.word;
        var fff = predictionsObj.Predict(ff);
        var isGood = fff.IsGood;

在我的培训数据中：

public static List<SuggestedFeedbackTrainingData> suggestedFeedbackTrainingData()
    {
        List<SuggestedFeedbackTrainingData> sftd = new List<SuggestedFeedbackTrainingData>();

        sftd.Add(new SuggestedFeedbackTrainingData()
        {
            SuggestedFeedBackText = "great",
            IsGood = true
        });

        sftd.Add(new SuggestedFeedbackTrainingData()
        {
            SuggestedFeedBackText = "good",
            IsGood = true
        });

        sftd.Add(new SuggestedFeedbackTrainingData()
        {
            SuggestedFeedBackText = "bad",
            IsGood = false
        });

        sftd.Add(new SuggestedFeedbackTrainingData()
        {
            SuggestedFeedBackText = "nice",
            IsGood = true
        });

        return sftd;
    }

当我的输入是“这是好的”时，我一直将属性"IsGood“false的值返回"true”。

我是不是遗漏了什么？

Answer 1

有多种因素可能会影响您的模型的性能。下面是一些关于如何改进模型的指导。通常，这涉及使用更多(和代表性数据)，尝试不同的算法，使用更多的列，等等。

https://learn.microsoft.com/dotnet/machine-learning/resources/improve-machine-learning-model-ml-net

因为这听起来像是刚刚开始，我建议在Visual中使用Model。它使用自动机器学习(，AutoML)帮助查找数据的“最佳”模型。不过，我要指出的是，AutoML不会帮助您克服没有足够代表性数据的问题。

https://dotnet.microsoft.com/learn/ml-dotnet/get-started-tutorial/data