ML.NET polynomial regression

Dockerfile

FROM jmacivor/dotnet-binder:0.1.1

ARG NB_USER=jovyan
ARG NB_UID=1000

USER $NB_USER

ENV HOME=/home/$NB_USER

WORKDIR $HOME

COPY mlnet_polynomial_regression.ipynb $HOME/mlnet_polynomial_regression.ipynb
COPY Position_Salaries.csv $HOME/Position_Salaries.csv

mlnet_polynomial_regression.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Polynomial Regression"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Using Statements, Classes and Loading the Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "source": [
    "#r \"nuget:Microsoft.ML,1.5.0\"\n",
    "#r \"nuget:Microsoft.ML.FastTree,1.5.0\"\n",
    "#r \"nuget:Microsoft.ML.Mkl.Components,1.5.0\"\n",
    "//Install XPlot package\n",
    "#r \"nuget:XPlot.Plotly,2.0.0\"   \n",
    "\n",
    "using System;\n",
    "using System.Collections.Generic;\n",
    "using System.Linq;\n",
    "using System.Composition;\n",
    "using Microsoft.ML;\n",
    "using Microsoft.ML.Data;\n",
    "using Microsoft.ML.Trainers;\n",
    "using Microsoft.ML.Transforms;\n",
    "using XPlot.Plotly;"
   ],
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": "Installed package XPlot.Plotly version 2.0.0"
     },
     "execution_count": 1,
     "metadata": {}
    },
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": "Installed package Microsoft.ML.FastTree version 1.5.0"
     },
     "execution_count": 1,
     "metadata": {}
    },
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": "Installed package Microsoft.ML version 1.5.0"
     },
     "execution_count": 1,
     "metadata": {}
    },
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": "Installed package Microsoft.ML.Mkl.Components version 1.5.0"
     },
     "execution_count": 1,
     "metadata": {}
    }
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "source": [
    "public class ModelInput\n",
    "{\n",
    "\n",
    "    [ColumnName(\"Position\"), LoadColumn(0)]\n",
    "    public string Position { get; set; }\n",
    "\n",
    "    [ColumnName(\"Level\"), LoadColumn(1)]\n",
    "    public float Level { get; set; }\n",
    "\n",
    "    [ColumnName(\"Salary\"), LoadColumn(2)]\n",
    "    public float Salary { get; set; }\n",
    "}\n",
    "\n",
    "public class ModelOutput\n",
    "{\n",
    "    public float Score { get; set; }\n",
    "}"
   ],
   "outputs": []
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "source": [
    "var mlContext = new MLContext();\n",
    "IDataView trainingDataView = mlContext.Data.LoadFromTextFile<ModelInput>(\n",
    "    path: \"./Position_Salaries.csv\",\n",
    "    hasHeader: true,\n",
    "    separatorChar: ',',\n",
    "    allowQuoting: true,\n",
    "    allowSparse: false);"
   ],
   "outputs": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Create the Pipeline"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Create the transform classes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "source": [
    "public class TransformOutput\n",
    "{\n",
    "    [VectorType(5)]\n",
    "    public float[] Features {get;set;} \n",
    "\n",
    "    public float Salary { get; set; }\n",
    "}"
   ],
   "outputs": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Create the custom transform"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "source": [
    "[CustomMappingFactoryAttribute(\"Features\")]\n",
    "public class PolynomialFeatures : CustomMappingFactory<ModelInput, TransformOutput>\n",
    "{\n",
    "\n",
    "    private readonly int _degree;\n",
    "    public PolynomialFeatures(int degree)\n",
    "    {\n",
    "        _degree = degree;\n",
    "    }\n",
    "\n",
    "    public void Transform(ModelInput input, TransformOutput output)\n",
    "    {\n",
    "        output.Features = Enumerable.Range(0, _degree + 1).Select(i => (float)(Math.Pow(input.Level, i))).ToArray();\n",
    "        output.Salary = input.Salary;\n",
    "    }\n",
    "\n",
    "    public override Action<ModelInput, TransformOutput> GetMapping()\n",
    "    {\n",
    "        return Transform;\n",
    "    } \n",
    "\n",
    "}\n",
    "\n",
    "var polyFeaturesTest = new PolynomialFeatures(2);\n",
    "var testInputs = Enumerable.Range(1, 10).Select(i => new ModelInput() {Level = i});\n",
    "var testOutputs = testInputs.Select(ti => \n",
    "{\n",
    "    var testOutput = new TransformOutput();\n",
    "    polyFeaturesTest.Transform(ti, testOutput);\n",
    "    return testOutput;\n",
    "});\n",
    "display(testOutputs)"
   ],
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/html": "<table><thead><tr><th><i>index</i></th><th>Features</th><th>Salary</th></tr></thead><tbody><tr><td>0</td><td><div class=\"dni-plaintext\">[ 1, 1, 1 ]</div></td><td><div class=\"dni-plaintext\">0</div></td></tr><tr><td>1</td><td><div class=\"dni-plaintext\">[ 1, 2, 4 ]</div></td><td><div class=\"dni-plaintext\">0</div></td></tr><tr><td>2</td><td><div class=\"dni-plaintext\">[ 1, 3, 9 ]</div></td><td><div class=\"dni-plaintext\">0</div></td></tr><tr><td>3</td><td><div class=\"dni-plaintext\">[ 1, 4, 16 ]</div></td><td><div class=\"dni-plaintext\">0</div></td></tr><tr><td>4</td><td><div class=\"dni-plaintext\">[ 1, 5, 25 ]</div></td><td><div class=\"dni-plaintext\">0</div></td></tr><tr><td>5</td><td><div class=\"dni-plaintext\">[ 1, 6, 36 ]</div></td><td><div class=\"dni-plaintext\">0</div></td></tr><tr><td>6</td><td><div class=\"dni-plaintext\">[ 1, 7, 49 ]</div></td><td><div class=\"dni-plaintext\">0</div></td></tr><tr><td>7</td><td><div class=\"dni-plaintext\">[ 1, 8, 64 ]</div></td><td><div class=\"dni-plaintext\">0</div></td></tr><tr><td>8</td><td><div class=\"dni-plaintext\">[ 1, 9, 81 ]</div></td><td><div class=\"dni-plaintext\">0</div></td></tr><tr><td>9</td><td><div class=\"dni-plaintext\">[ 1, 10, 100 ]</div></td><td><div class=\"dni-plaintext\">0</div></td></tr></tbody></table>"
     },
     "execution_count": 1,
     "metadata": {}
    }
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Create the pipeline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "source": [
    "// Data process configuration with pipeline data transformations \n",
    "public class CustomMappingOutput { public bool IsUnder5 {get;set;}}\n",
    "Action<ModelInput, CustomMappingOutput > mapping =\n",
    "                (input, output) => output.IsUnder5 = input.Level < 5;\n",
    "\n",
    "var polyFeatures = new PolynomialFeatures(4);\n",
    "var dataProcessPipeline = mlContext.Transforms.CustomMapping<ModelInput, TransformOutput>(polyFeatures.GetMapping(), contractName: \"PolynomialFeatures\")\n",
    "                            .Append(mlContext.Transforms.DropColumns(\"Position\", \"Level\"))\n",
    "                            .Append(mlContext.Transforms.Concatenate(\"Features\", new[] {\"Features\"}));\n",
    "\n",
    " var trainer = mlContext.Regression.Trainers.Ols(featureColumnName: \"Features\", labelColumnName: \"Salary\");\n",
    " var trainingPipeline = dataProcessPipeline.Append(trainer);"
   ],
   "outputs": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Train the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "source": [
    "Console.WriteLine(\"=============== Training  model ===============\");\n",
    "\n",
    "var transformedDataView = trainingPipeline.Fit(trainingDataView).Transform(trainingDataView);\n",
    "var transformedData = mlContext.Data.CreateEnumerable<TransformOutput>(transformedDataView, reuseRowObject: false);\n",
    "display(transformedData);\n",
    "var model = trainingPipeline.Fit(trainingDataView);\n",
    "\n",
    "Console.WriteLine(\"=============== End of training process ===============\");"
   ],
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": "=============== Training  model ===============\r\n"
     },
     "execution_count": 1,
     "metadata": {}
    },
    {
     "output_type": "execute_result",
     "data": {
      "text/html": "<table><thead><tr><th><i>index</i></th><th>Features</th><th>Salary</th></tr></thead><tbody><tr><td>0</td><td><div class=\"dni-plaintext\">[ 1, 1, 1, 1, 1 ]</div></td><td><div class=\"dni-plaintext\">45000</div></td></tr><tr><td>1</td><td><div class=\"dni-plaintext\">[ 1, 2, 4, 8, 16 ]</div></td><td><div class=\"dni-plaintext\">50000</div></td></tr><tr><td>2</td><td><div class=\"dni-plaintext\">[ 1, 3, 9, 27, 81 ]</div></td><td><div class=\"dni-plaintext\">60000</div></td></tr><tr><td>3</td><td><div class=\"dni-plaintext\">[ 1, 4, 16, 64, 256 ]</div></td><td><div class=\"dni-plaintext\">80000</div></td></tr><tr><td>4</td><td><div class=\"dni-plaintext\">[ 1, 5, 25, 125, 625 ]</div></td><td><div class=\"dni-plaintext\">110000</div></td></tr><tr><td>5</td><td><div class=\"dni-plaintext\">[ 1, 6, 36, 216, 1296 ]</div></td><td><div class=\"dni-plaintext\">150000</div></td></tr><tr><td>6</td><td><div class=\"dni-plaintext\">[ 1, 7, 49, 343, 2401 ]</div></td><td><div class=\"dni-plaintext\">200000</div></td></tr><tr><td>7</td><td><div class=\"dni-plaintext\">[ 1, 8, 64, 512, 4096 ]</div></td><td><div class=\"dni-plaintext\">300000</div></td></tr><tr><td>8</td><td><div class=\"dni-plaintext\">[ 1, 9, 81, 729, 6561 ]</div></td><td><div class=\"dni-plaintext\">500000</div></td></tr><tr><td>9</td><td><div class=\"dni-plaintext\">[ 1, 10, 100, 1000, 10000 ]</div></td><td><div class=\"dni-plaintext\">1000000</div></td></tr></tbody></table>"
     },
     "execution_count": 1,
     "metadata": {}
    },
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": "=============== End of training process ===============\r\n"
     },
     "execution_count": 1,
     "metadata": {}
    }
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Evaluate Trainer\n",
    "This is plain garbage for polynomial linear regression, sadly. But here it is anyway for your entertainment."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "source": [
    "var crossValidationResults = mlContext.Regression.CrossValidate(trainingDataView, trainingPipeline, numberOfFolds: 5, labelColumnName: \"Salary\");\n",
    "var L1 = crossValidationResults.Select(r => r.Metrics.MeanAbsoluteError);\n",
    "var L2 = crossValidationResults.Select(r => r.Metrics.MeanSquaredError);\n",
    "var RMS = crossValidationResults.Select(r => r.Metrics.RootMeanSquaredError);\n",
    "var lossFunction = crossValidationResults.Select(r => r.Metrics.LossFunction);\n",
    "var R2 = crossValidationResults.Select(r => r.Metrics.RSquared);\n",
    "\n",
    "Console.WriteLine($\"*************************************************************************************************************\");\n",
    "Console.WriteLine($\"*       Metrics for Regression model      \");\n",
    "Console.WriteLine($\"*------------------------------------------------------------------------------------------------------------\");\n",
    "Console.WriteLine($\"*       Average L1 Loss:       {L1.Average():0.###} \");\n",
    "Console.WriteLine($\"*       Average L2 Loss:       {L2.Average():0.###}  \");\n",
    "Console.WriteLine($\"*       Average RMS:           {RMS.Average():0.###}  \");\n",
    "Console.WriteLine($\"*       Average Loss Function: {lossFunction.Average():0.###}  \");\n",
    "Console.WriteLine($\"*       Average R-squared:     {R2.Average():0.###}  \");\n",
    "Console.WriteLine($\"*************************************************************************************************************\");\n",
    ""
   ],
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": "*************************************************************************************************************\r\n"
     },
     "execution_count": 1,
     "metadata": {}
    },
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": "*       Metrics for Regression model      \r\n"
     },
     "execution_count": 1,
     "metadata": {}
    },
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": "*------------------------------------------------------------------------------------------------------------\r\n"
     },
     "execution_count": 1,
     "metadata": {}
    },
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": "*       Average L1 Loss:       28752.58 \r\n"
     },
     "execution_count": 1,
     "metadata": {}
    },
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": "*       Average L2 Loss:       2536167340.344  \r\n"
     },
     "execution_count": 1,
     "metadata": {}
    },
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": "*       Average RMS:           33389.23  \r\n"
     },
     "execution_count": 1,
     "metadata": {}
    },
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": "*       Average Loss Function: 2536167335.8  \r\n"
     },
     "execution_count": 1,
     "metadata": {}
    },
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": "*       Average R-squared:     -∞  \r\n"
     },
     "execution_count": 1,
     "metadata": {}
    },
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": "*************************************************************************************************************\r\n"
     },
     "execution_count": 1,
     "metadata": {}
    }
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Show some results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "source": [
    "//predict\n",
    "var predEngine = mlContext.Model.CreatePredictionEngine<ModelInput, ModelOutput>(model);\n",
    "var predModel = new ModelInput() { Level=10f};\n",
    "display(predEngine.Predict(predModel));\n",
    "\n",
    "var predictions = Enumerable.Range(1,11).Select(i => predEngine.Predict(new ModelInput() { Level = i})).Select(o => o.Score);\n",
    "var actualSalaries = transformedData.Select(td => td.Salary);\n",
    "\n",
    "var actual = new Graph.Scatter()\n",
    "{\n",
    "    x = Enumerable.Range(1,11).ToArray(),\n",
    "    y = actualSalaries.ToArray(),\n",
    "    mode = \"markers\",\n",
    "    name = \"Actual\"\n",
    "};\n",
    "\n",
    "var predicted = new Graph.Scatter()\n",
    "{\n",
    "    x = Enumerable.Range(1,11).ToArray(),\n",
    "    y = predictions.ToArray(),\n",
    "    mode = \"line\",\n",
    "    name = \"Predicted\"\n",
    "};\n",
    "\n",
    "var chart = Chart.Plot(new[] {actual, predicted});\n",
    "var layout = new Layout.Layout(){barmode = \"group\", title=\"\"};\n",
    "chart.WithLayout(layout);\n",
    "chart.WithXTitle(\"Level\");\n",
    "chart.WithYTitle(\"Salary\");\n",
    "chart.WithLegend(true);\n",
    "chart.Width = 700;\n",
    "chart.Height = 400;\n",
    "\n",
    "display(chart);"
   ],
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/html": "<table><thead><tr><th>Score</th></tr></thead><tbody><tr><td><div class=\"dni-plaintext\">988917.75</div></td></tr></tbody></table>"
     },
     "execution_count": 1,
     "metadata": {}
    },
    {
     "output_type": "execute_result",
     "data": {
      "text/html": "<div id=\"e026095d-4c9b-441e-9de1-0ac943dc669d\" style=\"width: 700px; height: 400px;\"></div>\r\n<script type=\"text/javascript\">\r\n\r\nvar renderPlotly = function() {\r\n    var xplotRequire = require.config({context:'xplot-3.0.1',paths:{plotly:'https://cdn.plot.ly/plotly-1.49.2.min'}}) || require;\r\n    xplotRequire(['plotly'], function(Plotly) {\r\n\n            var data = [{\"type\":\"scatter\",\"x\":[1,2,3,4,5,6,7,8,9,10,11],\"y\":[45000.0,50000.0,60000.0,80000.0,110000.0,150000.0,200000.0,300000.0,500000.0,1000000.0],\"mode\":\"markers\",\"name\":\"Actual\"},{\"type\":\"scatter\",\"x\":[1,2,3,4,5,6,7,8,9,10,11],\"y\":[53358.9,31757.156,58640.023,94633.22,121727.22,143277.1,184003.72,289992.22,528692.25,988917.75,1780850.8],\"mode\":\"line\",\"name\":\"Predicted\"}];\n            var layout = {\"title\":\"\",\"showlegend\":true,\"xaxis\":{\"title\":\"Level\",\"_isSubplotObj\":true},\"yaxis\":{\"title\":\"Salary\",\"_isSubplotObj\":true},\"barmode\":\"group\"};\n            Plotly.newPlot('e026095d-4c9b-441e-9de1-0ac943dc669d', data, layout);\n        \r\n});\r\n};\r\n// ensure `require` is available globally\r\nif ((typeof(require) !==  typeof(Function)) || (typeof(require.config) !== typeof(Function))) {\r\n    let require_script = document.createElement('script');\r\n    require_script.setAttribute('src', 'https://cdnjs.cloudflare.com/ajax/libs/require.js/2.3.6/require.min.js');\r\n    require_script.setAttribute('type', 'text/javascript');\r\n    \r\n    \r\n    require_script.onload = function() {\r\n        renderPlotly();\r\n    };\r\n\r\n    document.getElementsByTagName('head')[0].appendChild(require_script);\r\n}\r\nelse {\r\n    renderPlotly();\r\n}\r\n\r\n</script>\r\n"
     },
     "execution_count": 1,
     "metadata": {}
    }
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": ".NET (C#)",
   "language": "C#",
   "name": ".net-csharp"
  },
  "language_info": {
   "file_extension": ".cs",
   "mimetype": "text/x-csharp",
   "name": "C#",
   "pygments_lexer": "csharp",
   "version": "8.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}

Position_Salaries.csv