aswi2021tri-musketyri-gitlab/Server/ServerApp/Predictor/NaiveBayesClassifier.cs @ cb0b7cb9

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Threading.Tasks;
using Microsoft.ML;
using Microsoft.ML.Data;
using ServerApp.Parser.OutputInfo;

namespace ServerApp.Predictor
{
    class NaiveBayesClassifier : IPredictor
    {
        private MLContext mlContext;

        private ITransformer model;

        public NaiveBayesClassifier()
        {
            mlContext = new MLContext();

        }

        public IEnumerable<ModelInput> ExtractModelInput(List<WeatherInfo> weatherInfos, List<ActivityInfo> activityInfos)
        {
            return weatherInfos.Select(e => new ModelInput(){
                Temp = (float)e.temp,
                Label = e.temp > 15.0 ? "Full" : "Empty",
            }).ToList();
        }

        public void Fit(IEnumerable<ModelInput> trainingData)
        {
            IDataView trainingDataView = mlContext.Data.LoadFromEnumerable(trainingData);
            var dataProcessPipeline = mlContext.Transforms.Conversion.MapValueToKey(nameof(ModelInput.Label))
                .Append(mlContext.Transforms.Concatenate("Features", new[] { "temp" })
                .Append(mlContext.Transforms.NormalizeMinMax("Features", "Features")));
            var trainer = mlContext.MulticlassClassification.Trainers.NaiveBayes();
            var traininingPipeline = dataProcessPipeline.Append(trainer)
                .Append(mlContext.Transforms.Conversion.MapKeyToValue("prediction", "PredictedLabel"));

            this.model = traininingPipeline.Fit(trainingDataView);

        }

        public String[] Predict(IEnumerable<ModelInput> input)
        {
            var data = mlContext.Data.LoadFromEnumerable(input);
            IDataView result = model.Transform(data);
            String[] prediction = result.GetColumn<String>("prediction").ToArray();

            return prediction;
        }
    }
}