AutoML on BBD Large Dataset

LargeDataset.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## Install AutoML NuGet package"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "dotnet_interactive": {
          "language": "csharp"
        }
      },
      "source": [
        "#i \"nuget:https://pkgs.dev.azure.com/dnceng/public/_packaging/MachineLearning/nuget/v3/index.json\"\n",
        "\n",
        "#r \"nuget:Microsoft.ML.AutoML,0.20.0-preview.22424.1\""
      ],
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": "<div><div><strong>Restore sources</strong><ul><li><span>https://pkgs.dev.azure.com/dnceng/public/_packaging/MachineLearning/nuget/v3/index.json</span></li></ul></div><div></div><div><strong>Installed Packages</strong><ul><li><span>Microsoft.ML.AutoML, 0.20.0-preview.22424.1</span></li></ul></div></div>"
          },
          "execution_count": 1,
          "metadata": {}
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/markdown": "Loading extensions from `Microsoft.ML.AutoML.Interactive.dll`"
          },
          "execution_count": 1,
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## Import packages"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "dotnet_interactive": {
          "language": "csharp"
        }
      },
      "source": [
        "using Microsoft.ML;\n",
        "using Microsoft.ML.Data;\n",
        "using Microsoft.ML.AutoML;"
      ],
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## Define input schema"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "dotnet_interactive": {
          "language": "csharp"
        }
      },
      "source": [
        "public class Input\n",
        "{\n",
        "\t[LoadColumn(0,149999)]\n",
        "\t[VectorType(150000)]\n",
        "\tpublic float[] Features {get;set;}\n",
        "\n",
        "\t[LoadColumn(150000)]\n",
        "\tpublic bool Label {get;set;}\n",
        "}"
      ],
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## Define data path"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "dotnet_interactive": {
          "language": "csharp"
        }
      },
      "source": [
        "var dataPath = @\"C:\\Datasets\\BBD_Full.csv\";"
      ],
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## Initialize MLContext"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "dotnet_interactive": {
          "language": "csharp"
        }
      },
      "source": [
        "var mlContext = new MLContext();"
      ],
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## Load data into IDataView"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "dotnet_interactive": {
          "language": "csharp"
        }
      },
      "source": [
        "var data = mlContext.Data.LoadFromTextFile<Input>(dataPath, hasHeader:true,separatorChar:',');"
      ],
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## Shuffle data\n",
        "\n",
        "**NOTE: This is not needed when using the full dataset. This is done so there's 0 and 1 values in the *Label* column since only 5k are sampled to simplify training in this sample.**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "dotnet_interactive": {
          "language": "csharp"
        }
      },
      "source": [
        "var shuffledData = mlContext.Data.ShuffleRows(data);"
      ],
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## Take 5k rows\n",
        "\n",
        "This is to make training faster. Not needed when using the entire dataset."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "dotnet_interactive": {
          "language": "csharp"
        }
      },
      "source": [
        "var sample = mlContext.Data.TakeRows(shuffledData,5000);"
      ],
      "outputs": []
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "dotnet_interactive": {
          "language": "csharp"
        }
      },
      "source": [
        "var trainTestData = mlContext.Data.TrainTestSplit(sample, testFraction:0.1);"
      ],
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## Display IDataView Schema"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "dotnet_interactive": {
          "language": "csharp"
        }
      },
      "source": [
        "data.Schema"
      ],
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": "<table><thead><tr><th><i>index</i></th><th>Name</th><th>Index</th><th>IsHidden</th><th>Type</th><th>Annotations</th></tr></thead><tbody><tr><td>0</td><td>Features</td><td><div class=\"dni-plaintext\">0</div></td><td><div class=\"dni-plaintext\">False</div></td><td><table><thead><tr><th>Dimensions</th><th>IsKnownSize</th><th>ItemType</th><th>Size</th><th>RawType</th></tr></thead><tbody><tr><td><div class=\"dni-plaintext\">[ 150000 ]</div></td><td><div class=\"dni-plaintext\">True</div></td><td><div class=\"dni-plaintext\">{ Single: RawType: System.Single }</div></td><td><div class=\"dni-plaintext\">150000</div></td><td><div class=\"dni-plaintext\">Microsoft.ML.Data.VBuffer&lt;System.Single&gt;</div></td></tr></tbody></table></td><td><table><thead><tr><th>Schema</th></tr></thead><tbody><tr><td><div class=\"dni-plaintext\">[  ]</div></td></tr></tbody></table></td></tr><tr><td>1</td><td>Label</td><td><div class=\"dni-plaintext\">1</div></td><td><div class=\"dni-plaintext\">False</div></td><td><table><thead><tr><th>RawType</th></tr></thead><tbody><tr><td><div class=\"dni-plaintext\">System.Boolean</div></td></tr></tbody></table></td><td><table><thead><tr><th>Schema</th></tr></thead><tbody><tr><td><div class=\"dni-plaintext\">[  ]</div></td></tr></tbody></table></td></tr></tbody></table>"
          },
          "execution_count": 1,
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## Define pipeline"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "dotnet_interactive": {
          "language": "csharp"
        }
      },
      "source": [
        "var pipeline = \n",
        "\tmlContext.Auto().Featurizer(trainTestData.TrainSet,numericColumns:new[] {\"Features\"})\n",
        "\t\t.Append(mlContext.Auto().BinaryClassification());"
      ],
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## Initialize AutoML experiment and configure settings"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "dotnet_interactive": {
          "language": "csharp"
        }
      },
      "source": [
        "var experiment = mlContext.Auto().CreateExperiment();\n",
        "\n",
        "experiment\n",
        "\t.SetPipeline(pipeline)\n",
        "\t.SetTrainingTimeInSeconds(60)\n",
        "\t.SetBinaryClassificationMetric(BinaryClassificationMetric.Accuracy, labelColumn:\"Label\")\n",
        "\t.SetDataset(trainTestData.TrainSet, trainTestData.TestSet);"
      ],
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## Run AutoML experiment"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "dotnet_interactive": {
          "language": "csharp"
        }
      },
      "source": [
        "var result = await experiment.RunAsync();"
      ],
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## Display metric"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "dotnet_interactive": {
          "language": "csharp"
        }
      },
      "source": [
        "result.Metric"
      ],
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": "<div class=\"dni-plaintext\">0.9866220735785953</div>"
          },
          "execution_count": 1,
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## Save model"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "dotnet_interactive": {
          "language": "csharp"
        }
      },
      "source": [
        "mlContext.Model.Save(result.Model, data.Schema, \"BBDModel.zip\");"
      ],
      "outputs": []
    }
  ],
  "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
}