ML.NET 算法正在评估有关乳腺癌的实例总是错误的

Question

我正在使用 ML.NET 二进制分类算法和威斯康星州乳腺癌数据。训练模型后，我发现每个实例都被评估为错误。在我的测试文件中，我有 100 个实例。75 个负数和 25 个正数。因此，从指标来看，准确度为 0.75，负精度为 0.75。这意味着所有实例都使用 0（假）进行评估。private static string trainingDataPath = Path.Combine(Path.Combine(AppDomain.CurrentDomain.BaseDirectory, "uploads"), "data.csv");        private static string testDataPath = Path.Combine(Path.Combine(AppDomain.CurrentDomain.BaseDirectory, "uploads"), "test.csv");        public bool checkDiagnostic (BreastCancerData input)        {             // set up a new machine learning context            var mlContext = new MLContext();            // load training and test data            var trainingDataView = mlContext.Data.LoadFromTextFile<BreastCancerData>(trainingDataPath, hasHeader: false, separatorChar: ',');            var testDataView = mlContext.Data.LoadFromTextFile<BreastCancerData>(testDataPath, hasHeader: false, separatorChar: ',');            // Preview the data.             //var dataPreview = trainingDataView.Preview(maxRows:700);            //var dataPreview2 = testDataView.Preview();            // the rest of the training code goes here...            var trainer = mlContext.BinaryClassification.Trainers.LinearSvm("Label", "Features");            var trainingPipeline = mlContext.Transforms.Concatenate(outputColumnName: "Features", nameof(BreastCancerData.AreaMean),                                nameof(BreastCancerData.AreaSe), nameof(BreastCancerData.AreaWorst), nameof(BreastCancerData.CompactnessMean),                                nameof(BreastCancerData.CompactnessSe), nameof(BreastCancerData.CompactnessWorst), nameof(BreastCancerData.ConcavePointsMean),                                nameof(BreastCancerData.ConcavePointsSe), nameof(BreastCancerData.ConcavePointsWorst), nameof(BreastCancerData.ConcavityMean),                                nameof(BreastCancerData.ConcavitySe), nameof(BreastCancerData.ConcavityWorst), nameof(BreastCancerData.FractalDimensionMean),

Answer 1

通过删除属性解决了问题Id。现在，SDCA 算法的准确率为 97.x%。