c# - 如何使用 ML.NET 预测多列

Question

我正在尝试创建一个应用程序，它根据用户的生活方式和药物的限制来预测服药的时间。
我的意思是:

我从一个病人那里得到如下信息:
• 他/她吃饭的次数和时间
• 他/她什么时候醒来并休眠
• 他/她必须服用多少药片

从医学的限制:
• 是否应该空腹吃药
• 药物是否应随餐服用/不随餐服用
• 患者是否需要在进餐和服药之间休息(它还没有显示在下面的屏幕上)
• 等等

示例数据集:
https://ibb.co/Gvry945

我应该使用什么类型的模型/力学/算法来预测服药时间？回归是正确的吗？我需要预测 1,2,3,4 有时是 5 列。

我写了一个简单的代码基于:
https://docs.microsoft.com/pl-pl/dotnet/machine-learning/tutorials/predict-prices
How to predict multiple labels with ML.NET using regression task?

它工作正常，我可以预测超过 1 列。但是，我的问题仍然是空白单元格。当我试图从该数据中预测某些内容时，它总是显示错误的值，并且只有在所有单元格都完成后才能正常工作。

那么，我应该将我的数据集分散到更少的数据集(所有单元格都完整)吗？前任。:
https://ibb.co/m8HVPvb
当我只预测 TimeToTakeMedicine1

https://ibb.co/qNk9xQL
当我预测 TimeToTakeMedicine1 和 TimeToTakeMedicine2 时

https://ibb.co/GnRc1c0
当我预测 TimeToTakeMedicine1、TimeToTakeMedicine2、TimeToTakeMedicine3 等时。

有没有更简单更好的方法来解决这个问题？

预测 TimeToTakeMedicine1、TimeToTakeMedicine2、TimeToTakeMedicine3 的工作代码(为了简单起见，我去掉了 OnEmptyStomach、WithMeal 和 IsPossible)

using System;
using System.IO;
using Microsoft.ML;
using Microsoft.ML.Trainers;

namespace NextTry
{
    class Program
    {
        static readonly string _trainDataPath = Path.Combine(Environment.CurrentDirectory, "DataFolder", "DataForPredictT1T2T3.csv");


        static void Main(string[] args)
        {

            MLContext mlContext = new MLContext(seed: 0);
            var model = Train(mlContext, _trainDataPath);

            TestSinglePrediction(mlContext, model);


        }

        public static ITransformer Train(MLContext mlContext, string dataPath)
        {
            IDataView dataView = mlContext.Data.LoadFromTextFile<Medicine>(dataPath, hasHeader: true, separatorChar: ',');

            var pipelineForMeal1 = mlContext.Transforms.CopyColumns(outputColumnName: "Label", inputColumnName: "TimeToTakeMedicine1")      
                .Append(mlContext.Transforms.Concatenate("Features", "MealTime1", "MealTime2", "MealTime3", "MealCount", "ActivityHoursWakeUp",  "ActivityHoursSleep", "PillsCount"))
                .Append(mlContext.Regression.Trainers.FastTree())
                .Append(mlContext.Transforms.CopyColumns(outputColumnName: "timeToTakeMedicine1", inputColumnName: "Score"));


            var pipelineForMeal2 = mlContext.Transforms.CopyColumns(outputColumnName: "Label", inputColumnName: "TimeToTakeMedicine2")
                .Append(mlContext.Transforms.Concatenate("Features", "MealTime1", "MealTime2", "MealTime3", "MealCount", "ActivityHoursWakeUp", "ActivityHoursSleep", "PillsCount"))
                .Append(mlContext.Regression.Trainers.FastTree())
                .Append(mlContext.Transforms.CopyColumns(outputColumnName: "timeToTakeMedicine2", inputColumnName: "Score"));


            var pipelineForMeal3 = mlContext.Transforms.CopyColumns(outputColumnName: "Label", inputColumnName: "TimeToTakeMedicine3")
                .Append(mlContext.Transforms.Concatenate("Features", "MealTime1", "MealTime2", "MealTime3", "MealCount", "ActivityHoursWakeUp", "ActivityHoursSleep",  "PillsCount"))
                .Append(mlContext.Regression.Trainers.FastTree())
                .Append(mlContext.Transforms.CopyColumns(outputColumnName: "timeToTakeMedicine3", inputColumnName: "Score"));


            var model = pipelineForMeal1
                .Append(pipelineForMeal2)
                .Append(pipelineForMeal3)
                .Fit(dataView);
            return model;
        }


        private static void TestSinglePrediction(MLContext mlContext, ITransformer model)
        {
            var predictionFunction = mlContext.Model.CreatePredictionEngine<Medicine, MedicineTimeTakeMedicinePrediction>(model);
            var medicineSample = new Medicine()
            {
                MealTime1 = 6,
                MealTime2 = 12,
                MealTime3 = 22,     
                MealCount = 3,
                PillsCount = 3
            };
            var prediction = predictionFunction.Predict(medicineSample);


            Console.WriteLine($"Predicted TimeToTakePill: {prediction.TimeToTakeMedicine1:0.####} ");
            Console.WriteLine($"Predicted TimeToTakePill: {prediction.TimeToTakeMedicine2:0.####}");
            Console.WriteLine($"Predicted TimeToTakePill: {prediction.TimeToTakeMedicine3:0.####}");


            Console.ReadKey();
        }
    }
}

using System;
using System.Collections.Generic;
using System.Text;
using Microsoft.ML.Data;

namespace NextTry
{
    public class Medicine

    {
        [LoadColumn(0)]
        public float MealTime1 { get; set; }

        [LoadColumn(1)]
        public float MealTime2 { get; set; }

        [LoadColumn(2)]
        public float MealTime3 { get; set; }

        [LoadColumn(3)]
        public float MealCount { get; set; }

        [LoadColumn(4)]
        public float ActivityHoursWakeUp { get; set; }

        [LoadColumn(5)]
        public float ActivityHoursSleep { get; set; }

        [LoadColumn(6)]
        public float PillsCount { get; set; }

        [LoadColumn(7)]
        public float TimeToTakeMedicine1 { get; set; }

        [LoadColumn(8)]
        public float TimeToTakeMedicine2 { get; set; }

        [LoadColumn(9)]
        public float TimeToTakeMedicine3 { get; set; }




    }
    public class MedicineTimeTakeMedicinePrediction

    {
        [ColumnName("timeToTakeMedicine1")]
        public float TimeToTakeMedicine1 { get; set; }

        [ColumnName("timeToTakeMedicine2")]
        public float TimeToTakeMedicine2 { get; set; }

        [ColumnName("timeToTakeMedicine3")]
        public float TimeToTakeMedicine3 { get; set; }


    }
}

Answer 1

我遇到了同样的问题。您要做的一件事是立即将所有模型附加到一个管道中，因为您具有相同的功能。