07_pandas_ml.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# pandas_ml"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false,
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [],
   "source": [
    "import seaborn as sns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>sepal_length</th>\n",
       "      <th>sepal_width</th>\n",
       "      <th>petal_length</th>\n",
       "      <th>petal_width</th>\n",
       "      <th>species</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>5.1</td>\n",
       "      <td>3.5</td>\n",
       "      <td>1.4</td>\n",
       "      <td>0.2</td>\n",
       "      <td>setosa</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>4.9</td>\n",
       "      <td>3.0</td>\n",
       "      <td>1.4</td>\n",
       "      <td>0.2</td>\n",
       "      <td>setosa</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>4.7</td>\n",
       "      <td>3.2</td>\n",
       "      <td>1.3</td>\n",
       "      <td>0.2</td>\n",
       "      <td>setosa</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4.6</td>\n",
       "      <td>3.1</td>\n",
       "      <td>1.5</td>\n",
       "      <td>0.2</td>\n",
       "      <td>setosa</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5.0</td>\n",
       "      <td>3.6</td>\n",
       "      <td>1.4</td>\n",
       "      <td>0.2</td>\n",
       "      <td>setosa</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   sepal_length  sepal_width  petal_length  petal_width species\n",
       "0           5.1          3.5           1.4          0.2  setosa\n",
       "1           4.9          3.0           1.4          0.2  setosa\n",
       "2           4.7          3.2           1.3          0.2  setosa\n",
       "3           4.6          3.1           1.5          0.2  setosa\n",
       "4           5.0          3.6           1.4          0.2  setosa"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "iris =sns.load_dataset(\"iris\")\n",
    "iris.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "- 説明/目的変数を別オブジェクトとして指定しなきゃいけないので煩雑\n",
    "- sklearnは素晴らしいけどpandasが普及したのでnumpy basedの出力がストレスフル。\n",
    "- sklearn使ってるとサブモジュールのimportが超多いのでめんどい。\n",
    "\n",
    "そこで**pandas_ml**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "ぶっちゃけると作者の[blog](http://sinhrks.hatenablog.com/entry/2015/03/19/234031)や[silde](https://speakerdeck.com/sinhrks/pandas-plus-scikit-learn-equals-pandas-ml-at-pydata-dot-tokyo-meetup-number-6-lightning-talk)を見た方がいいです。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true,
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [],
   "source": [
    "import pandas_ml as pdml"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>sepal_length</th>\n",
       "      <th>sepal_width</th>\n",
       "      <th>petal_length</th>\n",
       "      <th>petal_width</th>\n",
       "      <th>species</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>5.1</td>\n",
       "      <td>3.5</td>\n",
       "      <td>1.4</td>\n",
       "      <td>0.2</td>\n",
       "      <td>setosa</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>4.9</td>\n",
       "      <td>3.0</td>\n",
       "      <td>1.4</td>\n",
       "      <td>0.2</td>\n",
       "      <td>setosa</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>4.7</td>\n",
       "      <td>3.2</td>\n",
       "      <td>1.3</td>\n",
       "      <td>0.2</td>\n",
       "      <td>setosa</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4.6</td>\n",
       "      <td>3.1</td>\n",
       "      <td>1.5</td>\n",
       "      <td>0.2</td>\n",
       "      <td>setosa</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5.0</td>\n",
       "      <td>3.6</td>\n",
       "      <td>1.4</td>\n",
       "      <td>0.2</td>\n",
       "      <td>setosa</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   sepal_length  sepal_width  petal_length  petal_width species\n",
       "0           5.1          3.5           1.4          0.2  setosa\n",
       "1           4.9          3.0           1.4          0.2  setosa\n",
       "2           4.7          3.2           1.3          0.2  setosa\n",
       "3           4.6          3.1           1.5          0.2  setosa\n",
       "4           5.0          3.6           1.4          0.2  setosa"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pdml.ModelFrame(iris,target=\"species\")\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "ModelFrameってクラスに突っ込みます。見た目は変わらないです。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false,
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "iris:  <class 'pandas.core.frame.DataFrame'>\n",
      "df:  <class 'pandas_ml.core.frame.ModelFrame'>\n"
     ]
    }
   ],
   "source": [
    "print(\"iris: \",type(iris))\n",
    "print(\"df: \",type(df))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false,
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'_tree', 'predict_log_proba', '_decomposition', 'decision', '_wrap_probability', 'lm', 'multiclass', 'target', 'fit', '_cluster', 'data', 'target_name', 'pp', 'log_proba', '_semi_supervised', '_neighbors', 'neural_network', '_cross_validation', '_wrap_predicted', '_isotonic', 'feature_extraction', 'linear_model', '_mixture', 'ensemble', '_call', '_qda', 'crv', 'tree', '_DATA_NAME', '_concat_target', '_gaussian_process', 'mixture', 'has_target', 'decomposition', '_learning_curve', 'fit_predict', 'predicted', 'inverse_transform', 'random_projection', 'predict_proba', 'svm', 'preprocessing', 'cluster', '_naive_bayes', '_manifold', 'metrics', 'learning_curve', '_maybe_convert_data', 'predict', '_get_method_mapper', '_internal_caches', 'manifold', '_svm', '_grid_search', '_kernel_approximation', 'has_multi_targets', 'cls', '_covariance', 'proba', '_ensemble', 'transform', '_dummy', '_preprocessing', 'naive_bayes', 'grid_search', 'cross_validation', '_linear_model', '_pipeline', 'xgboost', 'xgb', '_random_projection', 'lda', 'dummy', '_metrics', '_cross_decomposition', 'covariance', 'neighbors', '_neural_network', '_feature_extraction', 'pipeline', '_multiclass', 'estimator', '_check_attr', 'semi_supervised', '_lda', '_feature_selection', '_xgboost', '_data_columns', '_method_mapper', 'gaussian_process', 'feature_selection', 'cross_decomposition', 'qda', 'fit_transform', 'decision_function', '_TARGET_NAME', 'isotonic', '_wrap_transform', 'score', 'kernel_approximation', 'has_data'}\n"
     ]
    }
   ],
   "source": [
    "print (set(dir(df)) - set(dir(iris)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "色々メソッドが増えてます。  \n",
    "基本的にsklearnのサブモジュールのものなので適当に探してください。  \n",
    "Jupyterなら補完が効くので楽です。  "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "## ポイント1\n",
    "ModelFrameにtargetを指定してあげると、説明変数と目的変数をメタデータとして持たせられる。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false,
    "slideshow": {
     "slide_type": "-"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>sepal_length</th>\n",
       "      <th>sepal_width</th>\n",
       "      <th>petal_length</th>\n",
       "      <th>petal_width</th>\n",
       "      <th>species</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>5.1</td>\n",
       "      <td>3.5</td>\n",
       "      <td>1.4</td>\n",
       "      <td>0.2</td>\n",
       "      <td>setosa</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>4.9</td>\n",
       "      <td>3.0</td>\n",
       "      <td>1.4</td>\n",
       "      <td>0.2</td>\n",
       "      <td>setosa</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>4.7</td>\n",
       "      <td>3.2</td>\n",
       "      <td>1.3</td>\n",
       "      <td>0.2</td>\n",
       "      <td>setosa</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4.6</td>\n",
       "      <td>3.1</td>\n",
       "      <td>1.5</td>\n",
       "      <td>0.2</td>\n",
       "      <td>setosa</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5.0</td>\n",
       "      <td>3.6</td>\n",
       "      <td>1.4</td>\n",
       "      <td>0.2</td>\n",
       "      <td>setosa</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   sepal_length  sepal_width  petal_length  petal_width species\n",
       "0           5.1          3.5           1.4          0.2  setosa\n",
       "1           4.9          3.0           1.4          0.2  setosa\n",
       "2           4.7          3.2           1.3          0.2  setosa\n",
       "3           4.6          3.1           1.5          0.2  setosa\n",
       "4           5.0          3.6           1.4          0.2  setosa"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false,
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>sepal_length</th>\n",
       "      <th>sepal_width</th>\n",
       "      <th>petal_length</th>\n",
       "      <th>petal_width</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>5.1</td>\n",
       "      <td>3.5</td>\n",
       "      <td>1.4</td>\n",
       "      <td>0.2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>4.9</td>\n",
       "      <td>3.0</td>\n",
       "      <td>1.4</td>\n",
       "      <td>0.2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>4.7</td>\n",
       "      <td>3.2</td>\n",
       "      <td>1.3</td>\n",
       "      <td>0.2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4.6</td>\n",
       "      <td>3.1</td>\n",
       "      <td>1.5</td>\n",
       "      <td>0.2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5.0</td>\n",
       "      <td>3.6</td>\n",
       "      <td>1.4</td>\n",
       "      <td>0.2</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   sepal_length  sepal_width  petal_length  petal_width\n",
       "0           5.1          3.5           1.4          0.2\n",
       "1           4.9          3.0           1.4          0.2\n",
       "2           4.7          3.2           1.3          0.2\n",
       "3           4.6          3.1           1.5          0.2\n",
       "4           5.0          3.6           1.4          0.2"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0    setosa\n",
       "1    setosa\n",
       "2    setosa\n",
       "3    setosa\n",
       "4    setosa\n",
       "Name: species, dtype: object"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.target.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "## ポイント2\n",
    "サブモジュールを呼ばなくてよい。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "頻出する機能はpandas_mlの内部でサブパッケージを呼んでくれるので、自分でimport書かなくてよい。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false,
    "scrolled": true,
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>species</th>\n",
       "      <th>sepal_length</th>\n",
       "      <th>sepal_width</th>\n",
       "      <th>petal_length</th>\n",
       "      <th>petal_width</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>setosa</td>\n",
       "      <td>0.803773</td>\n",
       "      <td>0.551609</td>\n",
       "      <td>0.220644</td>\n",
       "      <td>0.031521</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>setosa</td>\n",
       "      <td>0.828133</td>\n",
       "      <td>0.507020</td>\n",
       "      <td>0.236609</td>\n",
       "      <td>0.033801</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>setosa</td>\n",
       "      <td>0.805333</td>\n",
       "      <td>0.548312</td>\n",
       "      <td>0.222752</td>\n",
       "      <td>0.034269</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>setosa</td>\n",
       "      <td>0.800030</td>\n",
       "      <td>0.539151</td>\n",
       "      <td>0.260879</td>\n",
       "      <td>0.034784</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>setosa</td>\n",
       "      <td>0.790965</td>\n",
       "      <td>0.569495</td>\n",
       "      <td>0.221470</td>\n",
       "      <td>0.031639</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  species  sepal_length  sepal_width  petal_length  petal_width\n",
       "0  setosa      0.803773     0.551609      0.220644     0.031521\n",
       "1  setosa      0.828133     0.507020      0.236609     0.033801\n",
       "2  setosa      0.805333     0.548312      0.222752     0.034269\n",
       "3  setosa      0.800030     0.539151      0.260879     0.034784\n",
       "4  setosa      0.790965     0.569495      0.221470     0.031639"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.data=df.data.preprocessing.normalize()\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "モデルですら大体よべる。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "SVC(C=1.0, cache_size=200, class_weight=None, coef0=0.0,\n",
       "  decision_function_shape=None, degree=3, gamma='auto', kernel='rbf',\n",
       "  max_iter=-1, probability=False, random_state=None, shrinking=True,\n",
       "  tol=0.001, verbose=False)"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clf=df.svm.SVC()\n",
    "clf"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "ModelFrameにfitやpredict,transformなども実装されているので、df側からモデルを指定して各種実行可能。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "SVC(C=1.0, cache_size=200, class_weight=None, coef0=0.0,\n",
       "  decision_function_shape=None, degree=3, gamma='auto', kernel='rbf',\n",
       "  max_iter=-1, probability=False, random_state=None, shrinking=True,\n",
       "  tol=0.001, verbose=False)"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[:-10].fit(clf)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "140    versicolor\n",
       "141    versicolor\n",
       "142    versicolor\n",
       "143    versicolor\n",
       "144    versicolor\n",
       "145    versicolor\n",
       "146    versicolor\n",
       "147    versicolor\n",
       "148    versicolor\n",
       "149    versicolor\n",
       "dtype: object"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[-10:].predict(clf)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "全部が実装されているわけではないので、utils.shuffleなどは使えないが、pandasの機能でなんとかなる事が多い。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>species</th>\n",
       "      <th>sepal_length</th>\n",
       "      <th>sepal_width</th>\n",
       "      <th>petal_length</th>\n",
       "      <th>petal_width</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>setosa</td>\n",
       "      <td>0.817689</td>\n",
       "      <td>0.517314</td>\n",
       "      <td>0.250313</td>\n",
       "      <td>0.033375</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>virginica</td>\n",
       "      <td>0.724153</td>\n",
       "      <td>0.325344</td>\n",
       "      <td>0.566728</td>\n",
       "      <td>0.220394</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>versicolor</td>\n",
       "      <td>0.764350</td>\n",
       "      <td>0.355818</td>\n",
       "      <td>0.513959</td>\n",
       "      <td>0.158141</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>versicolor</td>\n",
       "      <td>0.740886</td>\n",
       "      <td>0.331740</td>\n",
       "      <td>0.552900</td>\n",
       "      <td>0.187986</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>setosa</td>\n",
       "      <td>0.784175</td>\n",
       "      <td>0.566349</td>\n",
       "      <td>0.246870</td>\n",
       "      <td>0.058087</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      species  sepal_length  sepal_width  petal_length  petal_width\n",
       "0      setosa      0.817689     0.517314      0.250313     0.033375\n",
       "1   virginica      0.724153     0.325344      0.566728     0.220394\n",
       "2  versicolor      0.764350     0.355818      0.513959     0.158141\n",
       "3  versicolor      0.740886     0.331740      0.552900     0.187986\n",
       "4      setosa      0.784175     0.566349      0.246870     0.058087"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.sample(frac=1).reset_index(drop=True).head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "fitで渡すインスタンスは別にpandas_mlから呼び出すものに限らず、sklearn対応ならなんでもよい。  \n",
    "サードパーティすらOK。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false,
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [],
   "source": [
    "from tensorflow.contrib import learn as skflow"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "DNN=skflow.TensorFlowDNNClassifier(n_classes=3,hidden_units=[100,100,100],steps=1000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#DNNはtargetを数値でしか受け付けないので変換しておく\n",
    "df2=df.copy()\n",
    "df2.target=df2.target.map({\"setosa\":0,\"versicolor\":1,\"virginica\":2})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": false,
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Step #100, epoch #20, avg. train loss: 0.52047\n",
      "Step #200, epoch #40, avg. train loss: 0.25982\n",
      "Step #300, epoch #60, avg. train loss: 0.17636\n",
      "Step #400, epoch #80, avg. train loss: 0.12365\n",
      "Step #500, epoch #100, avg. train loss: 0.11646\n",
      "Step #600, epoch #120, avg. train loss: 0.09757\n",
      "Step #700, epoch #140, avg. train loss: 0.10262\n",
      "Step #800, epoch #160, avg. train loss: 0.09943\n",
      "Step #900, epoch #180, avg. train loss: 0.10087\n",
      "Step #1000, epoch #200, avg. train loss: 0.09907\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "TensorFlowDNNClassifier(batch_size=32, class_weight=None, clip_gradients=5.0,\n",
       "            config=None, continue_training=False, dropout=None,\n",
       "            hidden_units=[100, 100, 100], learning_rate=0.1, n_classes=3,\n",
       "            optimizer='Adagrad', steps=1000, verbose=1)"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df2[:-10].fit(DNN)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": false,
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "140    2\n",
       "141    2\n",
       "142    2\n",
       "143    2\n",
       "144    2\n",
       "145    2\n",
       "146    2\n",
       "147    2\n",
       "148    2\n",
       "149    2\n",
       "dtype: int64"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df2[-10:].predict(DNN)"
   ]
  }
 ],
 "metadata": {
  "celltoolbar": "Slideshow",
  "kernelspec": {
   "display_name": "Environment (tf)",
   "language": "",
   "name": "tf"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.4.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}