ML.NET：功能列“功能”的架构不匹配

Question

I'm trying to learn ML.NET/Get into Machine Learning, but I'm stuck at an issue.

My goal is to create a Trained Model that can be used to predict a city based on input.

This code:

var dataPath = "cities.csv";
var mlContext = new MLContext();
var loader = mlContext.Data.CreateTextLoader<CityData>(hasHeader: false, separatorChar: ',');

var data = loader.Load(dataPath);

string featuresColumnName = "Features";

var pipeline = mlContext.Transforms.Concatenate(featuresColumnName, "PostalCode", "CityName")
        .Append(mlContext.Clustering.Trainers.KMeans(featuresColumnName, clustersCount: 3));

var model = pipeline.Fit(data);

Which should take an CSV as input (Which contains a list of Cities (Column 0 = Postal Code, Column 1 = CityName), and then add these features to the pipeline, gives the following error:

Unhandled Exception: System.ArgumentOutOfRangeException: Schema mismatch for feature column 'Features': expected Vector<R4>, got Vector<Text>

On the "Fit"- function.

I've done a bit of digging on the GitHub Repo, but I can't seem to find a solution. I'm working from the Iris- example ( https://docs.microsoft.com/en-us/dotnet/machine-learning/tutorials/iris-clustering ) (Of course with my modifications)

Any ideas?

Answer 1

Using FeaturizeText to transform strings features into a float array ones

var pipeline = mlContext.Transforms
    .Text.FeaturizeText("PostalCodeF", "PostalCode")
    .Append(mlContext.Transforms.Text.FeaturizeText("CityNameF", "CityName"))
    .Append(mlContext.Transforms.Concatenate(featuresColumnName, "PostalCodeF", "CityNameF"))
    .Append(mlContext.Clustering.Trainers.KMeans(featuresColumnName, clustersCount: 3));

var model = pipeline.Fit(data);