ML.NET using F#

This blog post shows the F# code that works for ML.NET. The code below read a text file with salary data. Trainer the model and predict the Salary. You can watch the original video in the link below.

open System

open Microsoft.ML

open Microsoft.ML.Data

let trainFile = “./SalaryTrain.csv”

let testFile = “./SalaryTest.csv”

type SalaryData = {

    [<LoadColumn(0)>]

    YearsExperience: float32;

    [<LoadColumn(1)>]

    Salary: float32;

}

[<CLIMutable>]

type SalaryPrediction = {

    [<ColumnName(“Score”)>]

    PredictedSalary: float32;

[<EntryPoint>]

let main argv =

        let context = MLContext()

        let loadDataByPath path =

            context.Data.LoadFromTextFile<SalaryData>(path, hasHeader=true, separatorChar=‘,’)

        let trainData = loadDataByPath trainFile

        let testData = loadDataByPath testFile

        let pipeline =

            EstimatorChain()

                .Append(context.Transforms.Concatenate(“Features”, “YearsExperience”))

                .Append(context.Transforms.CopyColumns((“Label”, “Salary”)))

                .Append(context.Regression.Trainers.LbfgsPoissonRegression())

        printfn “Training model…”

        let model = pipeline.Fit trainData

        let predictions = model.Transform testData

        printfn “Evaluating model…”

        let metrics = context.Regression.Evaluate(predictions, “Label”, “Score”)

        printfn “RMS – %.2f” metrics.RootMeanSquaredError

        printfn “R^2 – %.2f” metrics.RSquared

        let predictionFunc = context.Model.CreatePredictionEngine<SalaryData, SalaryPrediction>(model)

        let salaryPrediction = {

            YearsExperience = 8.0f

            Salary = 0.0f

        }

        let prediction = predictionFunc.Predict(salaryPrediction)

        printfn “Prediction – %.2f” prediction.PredictedSalary

        Console.ReadLine() |> ignore

        0

You will going to get the result below when you run the code. Do play around with the value and model.

Code download for this article only: https://github.com/chanmmn/ByOtherDeveloper/tree/main/FSharpRegression?WT.mc_id=DP-MVP-36769

Other code download: https://github.com/jwood803/MLNetExamples?WT.mc_id=DP-MVP-36769

Video: https://www.youtube.com/watch?v=PVJMX-iD0no&t=4s?WT.mc_id=DP-MVP-36769