动态构建管道链

Question

使用取自 this example 的代码。

此代码有效：

var pipeline = mlContext.Transforms.CopyColumns(outputColumnName: "Label", inputColumnName: "FareAmount")
            .Append(mlContext.Transforms.Categorical.OneHotEncoding(outputColumnName: "VendorIdEncoded", inputColumnName: "VendorId"))
            .Append(mlContext.Transforms.Categorical.OneHotEncoding(outputColumnName: "RateCodeEncoded", inputColumnName: "RateCode"))
            .Append(mlContext.Transforms.Categorical.OneHotEncoding(outputColumnName: "PaymentTypeEncoded", inputColumnName: "PaymentType"))
            .Append(mlContext.Transforms.Concatenate("Features", "VendorIdEncoded", "RateCodeEncoded", "PassengerCount", "TripTime", "TripDistance", "PaymentTypeEncoded"))
            .Append(mlContext.Regression.Trainers.FastTree());

但我想像这样在运行时动态追加：

var pipeline = mlContext.Transforms.CopyColumns(outputColumnName: "Label", inputColumnName: "FareAmount");
pipeline.Append(mlContext.Transforms.Categorical.OneHotEncoding(outputColumnName: "VendorIdEncoded", inputColumnName: "VendorId"));
pipeline.Append(mlContext.Transforms.Categorical.OneHotEncoding(outputColumnName: "RateCodeEncoded", inputColumnName: "RateCode"));
pipeline.Append(mlContext.Transforms.Categorical.OneHotEncoding(outputColumnName: "PaymentTypeEncoded", inputColumnName: "PaymentType"));
pipeline.Append(mlContext.Transforms.Concatenate("Features", "VendorIdEncoded", "RateCodeEncoded", "PassengerCount", "TripTime", "TripDistance", "PaymentTypeEncoded"));
pipeline.Append(mlContext.Regression.Trainers.FastTree());

None 后的第一个命令生效。如何动态链接 Appends？

尝试时：

var pipeline = mlContext.Transforms.CopyColumns(outputColumnName: "Label", inputColumnName: "FareAmount");
pipeline = pipeline.Append(mlContext.Transforms.Categorical.OneHotEncoding(outputColumnName: "VendorIdEncoded", inputColumnName: "VendorId"));
pipeline = pipeline.Append(mlContext.Transforms.Categorical.OneHotEncoding(outputColumnName: "RateCodeEncoded", inputColumnName: "RateCode"));
pipeline = pipeline.Append(mlContext.Transforms.Categorical.OneHotEncoding(outputColumnName: "PaymentTypeEncoded", inputColumnName: "PaymentType"));
pipeline = pipeline.Append(mlContext.Transforms.Concatenate("Features", "VendorIdEncoded", "RateCodeEncoded", "PassengerCount", "TripTime", "TripDistance", "PaymentTypeEncoded"));
pipeline = pipeline.Append(mlContext.Regression.Trainers.FastTree());

我收到 "Cannot implicitly convert type 'Microsoft.ML.Data.EstimatorChain' to 'Microsoft.ML.Transforms.ColumnCopyingEstimator'" 个错误

尝试时

var pipeline = mlContext.Transforms.CopyColumns(outputColumnName: "Label", inputColumnName: "FareAmount");
var pipeline2 = pipeline.Append(mlContext.Transforms.Categorical.OneHotEncoding(outputColumnName: "VendorIdEncoded", inputColumnName: "VendorId"));
pipeline2 = pipeline.Append(mlContext.Transforms.Categorical.OneHotEncoding(outputColumnName: "RateCodeEncoded", inputColumnName: "RateCode"));
pipeline2 = pipeline.Append(mlContext.Transforms.Categorical.OneHotEncoding(outputColumnName: "PaymentTypeEncoded", inputColumnName: "PaymentType"));
var pipeline3 = pipeline2.Append(mlContext.Transforms.Concatenate("Features", "VendorIdEncoded", "RateCodeEncoded", "PassengerCount", "TripTime", "TripDistance", "PaymentTypeEncoded"));
var pipeline4 = pipeline3.Append(mlContext.Regression.Trainers.FastTree());
var model = pipeline4.Fit(dataView);

但是在最后一行我得到一个“System.ArgumentOutOfRangeException: 'Could not find input column 'VendorIdEncoded' 参数名称：inputSchema'" error

Answer 1

正如@Camilo Terevinto 在评论部分解释的那样，您需要将结果输入 pipeline:

var pipeline = mlContext.Transforms.CopyColumns(outputColumnName: "Label", inputColumnName: "FareAmount");
var pipeline2 = pipeline.Append(mlContext.Transforms.Categorical.OneHotEncoding(outputColumnName: "VendorIdEncoded", inputColumnName: "VendorId"));
pipeline2 = pipeline.Append(mlContext.Transforms.Categorical.OneHotEncoding(outputColumnName: "RateCodeEncoded", inputColumnName: "RateCode"));
pipeline2 = pipeline.Append(mlContext.Transforms.Categorical.OneHotEncoding(outputColumnName: "PaymentTypeEncoded", inputColumnName: "PaymentType"));
pipeline2 = pipeline.Append(mlContext.Transforms.Concatenate("Features", "VendorIdEncoded", "RateCodeEncoded", "PassengerCount", "TripTime", "TripDistance", "PaymentTypeEncoded"));
pipeline2 = pipeline.Append(mlContext.Regression.Trainers.FastTree());

最终答案是

var pipeline = mlContext.Transforms.CopyColumns(outputColumnName: "Label", inputColumnName: "FareAmount");
var pipeline2 = pipeline.Append(mlContext.Transforms.Categorical.OneHotEncoding(outputColumnName: "VendorIdEncoded", inputColumnName: "VendorId"));
pipeline2 = pipeline2.Append(mlContext.Transforms.Categorical.OneHotEncoding(outputColumnName: "RateCodeEncoded", inputColumnName: "RateCode"));
pipeline2 = pipeline2.Append(mlContext.Transforms.Categorical.OneHotEncoding(outputColumnName: "PaymentTypeEncoded", inputColumnName: "PaymentType"));
var pipeline3 = pipeline2.Append(mlContext.Transforms.Concatenate("Features", "VendorIdEncoded", "RateCodeEncoded", "PassengerCount", "TripTime", "TripDistance", "PaymentTypeEncoded"));
var pipeline4 = pipeline3.Append(mlContext.Regression.Trainers.FastTree());
var model = pipeline4.Fit(dataView);

动态构建管道链

Dynamically build pipeline chain

c#

ml.net