29. Cross Validation.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "intelligent-frequency",
   "metadata": {},
   "source": [
    "# Cross Validation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "tribal-publication",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from sklearn import datasets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "centered-albania",
   "metadata": {},
   "outputs": [],
   "source": [
    "digits = datasets.load_digits()\n",
    "X = digits.data\n",
    "y = digits.target"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "successful-caution",
   "metadata": {},
   "source": [
    "## Train_test_split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "stuffed-finder",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.4, random_state=666)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "textile-listing",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best K = 3\n",
      "Best P = 4\n",
      "Best score = 0.9860917941585535\n",
      "CPU times: user 19.1 s, sys: 251 ms, total: 19.4 s\n",
      "Wall time: 19.7 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "\n",
    "best_score, best_p, best_k = 0, 0, 0\n",
    "for k in range(2, 10):\n",
    "    for p in range(1, 6):\n",
    "        knn_clf = KNeighborsClassifier(weights='distance', n_neighbors=k, p=p)\n",
    "        knn_clf.fit(X_train, y_train)\n",
    "        score = knn_clf.score(X_test, y_test)\n",
    "        if score > best_score:\n",
    "            best_score, best_p, best_k = score, p, k\n",
    "            \n",
    "print(\"Best K =\", best_k)\n",
    "print(\"Best P =\", best_p)\n",
    "print(\"Best score =\", best_score)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "beautiful-florist",
   "metadata": {},
   "source": [
    "## Cross Validation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "animated-hampton",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.99537037, 0.98148148, 0.97685185, 0.97674419, 0.97209302])"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.model_selection import cross_val_score\n",
    "\n",
    "knn_clf = KNeighborsClassifier()\n",
    "cross_val_score(knn_clf, X_train, y_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "accredited-abraham",
   "metadata": {},
   "source": [
    "\n",
    "1.拟合方式：cross_val_score(knn_clf, X_train, y_train)，默认将 X_train 分割成 3 份，并得到 3 个模型的准确率；如果想将 X_train 分割成 k 份cross_val_score(knn_clf, X_train, y_train, cv=k)；\n",
    "\n",
    "2.判定条件：score = np.mean(scores)，交叉验证中取 3 个模型的准确率的平均值最高时对应的一组参数作为最终的最优超参数；"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "communist-tract",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best K = 2\n",
      "Best P = 2\n",
      "Best score = 0.9851507321274763\n",
      "CPU times: user 25.5 s, sys: 396 ms, total: 25.9 s\n",
      "Wall time: 26.6 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "best_score, best_p, best_k = 0, 0, 0\n",
    "for k in range(2, 10):\n",
    "    for p in range(1, 6):\n",
    "        knn_clf = KNeighborsClassifier(weights='distance', n_neighbors=k, p=p)\n",
    "        scores = cross_val_score(knn_clf, X_train, y_train)\n",
    "        score = np.mean(scores)\n",
    "        if score > best_score:\n",
    "            best_score, best_p, best_k = score, p, k\n",
    "\n",
    "print(\"Best K =\", best_k)\n",
    "print(\"Best P =\", best_p)\n",
    "print(\"Best score =\", best_score)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "positive-ownership",
   "metadata": {},
   "outputs": [],
   "source": [
    "best_knn_clf = KNeighborsClassifier(weights = 'distance', n_neighbors = 2, p = 2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "cultural-broad",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.980528511821975"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "best_knn_clf.fit(X_train, y_train)\n",
    "best_knn_clf.score(X_test, y_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "demographic-intermediate",
   "metadata": {},
   "source": [
    "## Grid Search"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "hindu-consortium",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 3 folds for each of 45 candidates, totalling 135 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Using backend SequentialBackend with 1 concurrent workers.\n",
      "[Parallel(n_jobs=1)]: Done 135 out of 135 | elapsed:   24.3s finished\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "GridSearchCV(cv=3,\n",
       "             estimator=KNeighborsClassifier(n_neighbors=9, p=5,\n",
       "                                            weights='distance'),\n",
       "             param_grid=[{'n_neighbors': [2, 3, 4, 5, 6, 7, 8, 9, 10],\n",
       "                          'p': [1, 2, 3, 4, 5], 'weights': ['distance']}],\n",
       "             verbose=1)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.model_selection import GridSearchCV\n",
    "\n",
    "param_grid = [\n",
    "    {\n",
    "        'weights': ['distance'],\n",
    "        'n_neighbors': [i for i in range(2, 11)],\n",
    "        'p': [i for i in range(1, 6)]\n",
    "    }\n",
    "]\n",
    "\n",
    "grid_search = GridSearchCV(knn_clf, param_grid, verbose=1, cv=3)\n",
    "grid_search.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "metallic-technology",
   "metadata": {},
   "source": [
    "\n",
    "1. “3 folds”：就是指网格搜索使用交叉验证的方式，默认将 X_train 分割成 3 份；如果将 X_train 分割成 k 份：grid_search = GridSearchCV(knn_clf, param_grid, verbose=1, cv=k)；\n",
    "2. “45 candidates”：k 从 [2, 11) 有 9 个值，p 从 [1, 6) 有 5 个值，一共要对 45 组参数进行搜索；\n",
    "\n",
    "3. “135 fits”：一共要对 45 组参数进行搜索，每次搜索要训练出 3 个模型，一共要进行 135 此拟合训练；"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "integral-leeds",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9833023831631073"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "grid_search.best_score_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "level-algorithm",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'n_neighbors': 2, 'p': 2, 'weights': 'distance'}"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "grid_search.best_params_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "every-closer",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.980528511821975"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "best_knn_clf = grid_search.best_estimator_\n",
    "best_knn_clf.score(X_test, y_test)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}