duolingo_pred.ipynb

duolingo_pred.ipynb

{
  "cells": [
    {
      "metadata": {
        "trusted": true,
        "collapsed": true
      },
      "cell_type": "code",
      "source": "%matplotlib inline",
      "execution_count": 137,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true,
        "collapsed": true
      },
      "cell_type": "code",
      "source": "import pandas as pd\nimport numpy as np\nimport scipy.stats\nimport random\nfrom sklearn import linear_model\nimport matplotlib.pyplot as plt\nfrom sklearn.metrics import mean_squared_error\nfrom sklearn.metrics import mean_absolute_error",
      "execution_count": 118,
      "outputs": []
    },
    {
      "metadata": {},
      "cell_type": "markdown",
      "source": "# Sample and preprocess data"
    },
    {
      "metadata": {
        "trusted": true,
        "collapsed": true
      },
      "cell_type": "code",
      "source": "learning_traces_path = \"./Studio/data/learning_traces.13m.csv\"",
      "execution_count": 119,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true,
        "collapsed": true
      },
      "cell_type": "code",
      "source": "# number of records in file (excludes header)\nn = sum(1 for line in open(learning_traces_path)) - 1 \ns = 100000 #desired sample size\n#the 0-indexed header will not be included in the skip list\nskip = sorted(random.sample(range(1,n+1),n-s)) ",
      "execution_count": 120,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true,
        "collapsed": true
      },
      "cell_type": "code",
      "source": "learning_traces = pd.read_csv(learning_traces_path, skiprows=skip)",
      "execution_count": 121,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "learning_traces.head(10)",
      "execution_count": 122,
      "outputs": [
        {
          "data": {
            "text/plain": "   p_recall   timestamp   delta user_id learning_language ui_language  \\\n0      1.00  1362082527     179  u:h5ex                de          en   \n1      1.00  1362082533     443  u:ezFh                es          en   \n2      1.00  1362082561     269  u:gZQD                en          es   \n3      1.00  1362082574     547   u:mKI                es          en   \n4      1.00  1362082598   51714  u:iISw                es          en   \n5      1.00  1362082603     133  u:hZ9T                es          en   \n6      0.75  1362082610    1151  u:gA8I                en          pt   \n7      1.00  1362082614  185547  u:idJd                fr          en   \n8      1.00  1362082652     438  u:ecYb                en          es   \n9      1.00  1362082655  197936   u:ubz                es          en   \n\n                          lexeme_id                     lexeme_string  \\\n0  9e1f56b08922d1d9f7ab663b58d88367  freundin/freundin<n><f><sg><acc>   \n1  99cd0848be9239250ffd99e11add7338                         de/de<pr>   \n2  7db6cb705cd5fecf3fa67ee708c8854b               between/between<pr>   \n3  07605081e461f8cc70d0e27a91d1ba37        tortugas/tortuga<n><f><pl>   \n4  d64b11ab44cc6fa1b32ce3d4998bd5af          pájaros/pájaro<n><m><pl>   \n5  36cc2f758a27a4ebeb9f23d82881a811        pescados/pescado<n><m><pl>   \n6  f254244207240db11107ffbe433d1e0e                  bird/bird<n><sg>   \n7  3a05062c9117beb722ae9acb73b53eb6            bonnes/bon<adj><f><pl>   \n8  384ecdedd0e8caea25f8ba4e2798bf2b                orange/orange<adj>   \n9  2a5f9687f146bdb703ab03f7bedf2d3c             arroz/arroz<n><m><sg>   \n\n   history_seen  history_correct  session_seen  session_correct  \n0             4                4             2                2  \n1           368              330             1                1  \n2            36               34             2                2  \n3             5                4             1                1  \n4             1                1             1                1  \n5             6                6             3                3  \n6             1                1             4                3  \n7             4                4             2                2  \n8             5                4             1                1  \n9            35               35             2                2  ",
            "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>p_recall</th>\n      <th>timestamp</th>\n      <th>delta</th>\n      <th>user_id</th>\n      <th>learning_language</th>\n      <th>ui_language</th>\n      <th>lexeme_id</th>\n      <th>lexeme_string</th>\n      <th>history_seen</th>\n      <th>history_correct</th>\n      <th>session_seen</th>\n      <th>session_correct</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>1.00</td>\n      <td>1362082527</td>\n      <td>179</td>\n      <td>u:h5ex</td>\n      <td>de</td>\n      <td>en</td>\n      <td>9e1f56b08922d1d9f7ab663b58d88367</td>\n      <td>freundin/freundin&lt;n&gt;&lt;f&gt;&lt;sg&gt;&lt;acc&gt;</td>\n      <td>4</td>\n      <td>4</td>\n      <td>2</td>\n      <td>2</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>1.00</td>\n      <td>1362082533</td>\n      <td>443</td>\n      <td>u:ezFh</td>\n      <td>es</td>\n      <td>en</td>\n      <td>99cd0848be9239250ffd99e11add7338</td>\n      <td>de/de&lt;pr&gt;</td>\n      <td>368</td>\n      <td>330</td>\n      <td>1</td>\n      <td>1</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>1.00</td>\n      <td>1362082561</td>\n      <td>269</td>\n      <td>u:gZQD</td>\n      <td>en</td>\n      <td>es</td>\n      <td>7db6cb705cd5fecf3fa67ee708c8854b</td>\n      <td>between/between&lt;pr&gt;</td>\n      <td>36</td>\n      <td>34</td>\n      <td>2</td>\n      <td>2</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>1.00</td>\n      <td>1362082574</td>\n      <td>547</td>\n      <td>u:mKI</td>\n      <td>es</td>\n      <td>en</td>\n      <td>07605081e461f8cc70d0e27a91d1ba37</td>\n      <td>tortugas/tortuga&lt;n&gt;&lt;f&gt;&lt;pl&gt;</td>\n      <td>5</td>\n      <td>4</td>\n      <td>1</td>\n      <td>1</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>1.00</td>\n      <td>1362082598</td>\n      <td>51714</td>\n      <td>u:iISw</td>\n      <td>es</td>\n      <td>en</td>\n      <td>d64b11ab44cc6fa1b32ce3d4998bd5af</td>\n      <td>pájaros/pájaro&lt;n&gt;&lt;m&gt;&lt;pl&gt;</td>\n      <td>1</td>\n      <td>1</td>\n      <td>1</td>\n      <td>1</td>\n    </tr>\n    <tr>\n      <th>5</th>\n      <td>1.00</td>\n      <td>1362082603</td>\n      <td>133</td>\n      <td>u:hZ9T</td>\n      <td>es</td>\n      <td>en</td>\n      <td>36cc2f758a27a4ebeb9f23d82881a811</td>\n      <td>pescados/pescado&lt;n&gt;&lt;m&gt;&lt;pl&gt;</td>\n      <td>6</td>\n      <td>6</td>\n      <td>3</td>\n      <td>3</td>\n    </tr>\n    <tr>\n      <th>6</th>\n      <td>0.75</td>\n      <td>1362082610</td>\n      <td>1151</td>\n      <td>u:gA8I</td>\n      <td>en</td>\n      <td>pt</td>\n      <td>f254244207240db11107ffbe433d1e0e</td>\n      <td>bird/bird&lt;n&gt;&lt;sg&gt;</td>\n      <td>1</td>\n      <td>1</td>\n      <td>4</td>\n      <td>3</td>\n    </tr>\n    <tr>\n      <th>7</th>\n      <td>1.00</td>\n      <td>1362082614</td>\n      <td>185547</td>\n      <td>u:idJd</td>\n      <td>fr</td>\n      <td>en</td>\n      <td>3a05062c9117beb722ae9acb73b53eb6</td>\n      <td>bonnes/bon&lt;adj&gt;&lt;f&gt;&lt;pl&gt;</td>\n      <td>4</td>\n      <td>4</td>\n      <td>2</td>\n      <td>2</td>\n    </tr>\n    <tr>\n      <th>8</th>\n      <td>1.00</td>\n      <td>1362082652</td>\n      <td>438</td>\n      <td>u:ecYb</td>\n      <td>en</td>\n      <td>es</td>\n      <td>384ecdedd0e8caea25f8ba4e2798bf2b</td>\n      <td>orange/orange&lt;adj&gt;</td>\n      <td>5</td>\n      <td>4</td>\n      <td>1</td>\n      <td>1</td>\n    </tr>\n    <tr>\n      <th>9</th>\n      <td>1.00</td>\n      <td>1362082655</td>\n      <td>197936</td>\n      <td>u:ubz</td>\n      <td>es</td>\n      <td>en</td>\n      <td>2a5f9687f146bdb703ab03f7bedf2d3c</td>\n      <td>arroz/arroz&lt;n&gt;&lt;m&gt;&lt;sg&gt;</td>\n      <td>35</td>\n      <td>35</td>\n      <td>2</td>\n      <td>2</td>\n    </tr>\n  </tbody>\n</table>\n</div>"
          },
          "output_type": "execute_result",
          "metadata": {},
          "execution_count": 122
        }
      ]
    },
    {
      "metadata": {
        "trusted": true,
        "collapsed": true
      },
      "cell_type": "code",
      "source": "data = learning_traces[[\"delta\", \"history_seen\", \"history_correct\", \"session_seen\", \"session_correct\", \"p_recall\"]].as_matrix()",
      "execution_count": 123,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true,
        "collapsed": true
      },
      "cell_type": "code",
      "source": "X = data[:, :5]\nY = data[:, 5:]",
      "execution_count": 124,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true,
        "collapsed": true
      },
      "cell_type": "code",
      "source": "train_test_ratio = 0.8\ntrain_size = int(np.floor(train_test_ratio * X.shape[0]))",
      "execution_count": 125,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "train_size",
      "execution_count": 136,
      "outputs": [
        {
          "data": {
            "text/plain": "80000"
          },
          "output_type": "execute_result",
          "metadata": {},
          "execution_count": 136
        }
      ]
    },
    {
      "metadata": {
        "trusted": true,
        "collapsed": true
      },
      "cell_type": "code",
      "source": "X_train, X_test = X[:train_size,:], X[train_size:,:]\nY_train, Y_test = Y[:train_size,:], Y[train_size:,:]",
      "execution_count": 126,
      "outputs": []
    },
    {
      "metadata": {},
      "cell_type": "markdown",
      "source": "# Predict with baseline constant"
    },
    {
      "metadata": {
        "trusted": true,
        "collapsed": true
      },
      "cell_type": "code",
      "source": "Y_train_predicted_constant, Y_test_predicted_constant = np.ones(Y_train.shape) * 0.859, np.ones(Y_test.shape) * 0.859",
      "execution_count": 127,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "mean_absolute_error(Y_train, Y_train_predicted_constant)",
      "execution_count": 128,
      "outputs": [
        {
          "data": {
            "text/plain": "0.19978840807296061"
          },
          "output_type": "execute_result",
          "metadata": {},
          "execution_count": 128
        }
      ]
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "mean_absolute_error(Y_test, Y_test_predicted_constant)",
      "execution_count": 129,
      "outputs": [
        {
          "data": {
            "text/plain": "0.19901131964983185"
          },
          "output_type": "execute_result",
          "metadata": {},
          "execution_count": 129
        }
      ]
    },
    {
      "metadata": {},
      "cell_type": "markdown",
      "source": "# Predict with linear regression"
    },
    {
      "metadata": {
        "trusted": true,
        "collapsed": true
      },
      "cell_type": "code",
      "source": "reg = linear_model.LinearRegression()",
      "execution_count": 130,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "reg.fit(X_train, Y_train)",
      "execution_count": 131,
      "outputs": [
        {
          "data": {
            "text/plain": "LinearRegression(copy_X=True, fit_intercept=True, n_jobs=1, normalize=False)"
          },
          "output_type": "execute_result",
          "metadata": {},
          "execution_count": 131
        }
      ]
    },
    {
      "metadata": {},
      "cell_type": "markdown",
      "source": "Predict and measure error on training data"
    },
    {
      "metadata": {
        "trusted": true,
        "collapsed": true
      },
      "cell_type": "code",
      "source": "Y_train_predicted_linear_regression = reg.predict(X_train)",
      "execution_count": 132,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "mean_absolute_error(Y_train, Y_train_predicted_linear_regression)",
      "execution_count": 133,
      "outputs": [
        {
          "data": {
            "text/plain": "0.08409331230676749"
          },
          "output_type": "execute_result",
          "metadata": {},
          "execution_count": 133
        }
      ]
    },
    {
      "metadata": {},
      "cell_type": "markdown",
      "source": "Predict and measure error on test data"
    },
    {
      "metadata": {
        "trusted": true,
        "collapsed": true
      },
      "cell_type": "code",
      "source": "Y_test_predicted_linear_regression = reg.predict(X_test)",
      "execution_count": 134,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true
      },
      "cell_type": "code",
      "source": "mean_absolute_error(Y_test, Y_test_predicted_linear_regression)",
      "execution_count": 135,
      "outputs": [
        {
          "data": {
            "text/plain": "0.084153068502124867"
          },
          "output_type": "execute_result",
          "metadata": {},
          "execution_count": 135
        }
      ]
    },
    {
      "metadata": {
        "trusted": true,
        "collapsed": true
      },
      "cell_type": "code",
      "source": "",
      "execution_count": null,
      "outputs": []
    }
  ],
  "metadata": {
    "language_info": {
      "file_extension": ".py",
      "codemirror_mode": {
        "version": 3,
        "name": "ipython"
      },
      "version": "3.5.4",
      "mimetype": "text/x-python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "name": "python"
    },
    "kernelspec": {
      "name": "conda-root-py",
      "display_name": "Python [conda root]",
      "language": "python"
    },
    "gist_info": {},
    "toc": {
      "toc_threshold": 6,
      "toc_number_sections": true,
      "toc_cell": false,
      "toc_window_display": false
    },
    "gist": {
      "id": "",
      "data": {
        "description": "duolingo_pred.ipynb",
        "public": false
      }
    }
  },
  "nbformat": 4,
  "nbformat_minor": 2
}