arvore_decisao.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Tree Decision Model Machine Learning Titanic"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      " Extraindo alvo e caracteristicas dos dados\n"
     ]
    }
   ],
   "source": [
    "import pandas as panda \n",
    "import limparDados\n",
    "from sklearn import tree, model_selection\n",
    "\n",
    "treinamento = panda.read_csv(\"train.csv\")\n",
    "teste = panda.read_csv(\"test.csv\")\n",
    "\n",
    "#Adicionando dados aos campos vazios\n",
    "limparDados.limpar_dados(treinamento)\n",
    "limparDados.limpar_dados(teste)\n",
    "print(\"\\n Extraindo alvo e caracteristicas dos dados\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.9797979797979798\n",
      "\n",
      " Realizando previsao no DataSet de Testes\n"
     ]
    }
   ],
   "source": [
    "alvo = treinamento[\"Survived\"].values\n",
    "#Escolhendo os features baseado na vizualização dos dados\n",
    "nomes_feat = [\"Pclass\", \"Age\", \"Fare\", \"Embarked\", \"Sex\", \"SibSp\", \"Parch\"]\n",
    "feat = treinamento[nomes_feat].values\n",
    "\n",
    "#Construindo a arvore de decisões\n",
    "arvore_decisao = tree.DecisionTreeClassifier(random_state=1)\n",
    "arvore_decisao = arvore_decisao.fit(feat, alvo)\n",
    "print(arvore_decisao.score(feat, alvo))\n",
    "\n",
    "print(\"\\n Realizando previsao no DataSet de Testes\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      " Teste realizado\n"
     ]
    }
   ],
   "source": [
    "teste_feat = teste[[\"Pclass\", \"Age\", \"Fare\", \"Embarked\", \"Sex\", \"SibSp\", \"Parch\"]].values\n",
    "#realizando a previsão de acordo com as features escolhidas\n",
    "previsao = arvore_decisao.predict(teste_feat)\n",
    "limparDados.escrever_previsao(previsao, \"arvore_decisao.csv\")\n",
    "\n",
    "print(\"\\n Teste realizado\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.17060767 0.12783353 0.15175864 0.01001273 0.47976487 0.05205183\n",
      " 0.00797074]\n",
      "0.8787878787878788\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[Text(140.10652173913044, 203.85, 'Sex <= 0.5\\nentropy = 0.473\\nsamples = 891\\nvalue = [549, 342]'),\n",
       " Text(47.61195652173913, 176.67000000000002, 'Age <= 6.5\\nentropy = 0.306\\nsamples = 577\\nvalue = [468, 109]'),\n",
       " Text(9.704347826086957, 149.49, 'SibSp <= 2.5\\nentropy = 0.444\\nsamples = 24\\nvalue = [8, 16]'),\n",
       " Text(4.852173913043479, 122.31, 'entropy = 0.0\\nsamples = 15\\nvalue = [0, 15]'),\n",
       " Text(14.556521739130435, 122.31, 'Age <= 2.5\\nentropy = 0.198\\nsamples = 9\\nvalue = [8, 1]'),\n",
       " Text(9.704347826086957, 95.13, 'entropy = 0.0\\nsamples = 5\\nvalue = [5, 0]'),\n",
       " Text(19.408695652173915, 95.13, 'Fare <= 31.331\\nentropy = 0.375\\nsamples = 4\\nvalue = [3, 1]'),\n",
       " Text(14.556521739130435, 67.94999999999999, 'entropy = 0.0\\nsamples = 3\\nvalue = [3, 0]'),\n",
       " Text(24.260869565217394, 67.94999999999999, 'entropy = 0.0\\nsamples = 1\\nvalue = [0, 1]'),\n",
       " Text(85.51956521739132, 149.49, 'Pclass <= 1.5\\nentropy = 0.28\\nsamples = 553\\nvalue = [460, 93]'),\n",
       " Text(46.095652173913045, 122.31, 'Fare <= 26.144\\nentropy = 0.46\\nsamples = 120\\nvalue = [77, 43]'),\n",
       " Text(41.243478260869566, 95.13, 'entropy = 0.0\\nsamples = 10\\nvalue = [10, 0]'),\n",
       " Text(50.947826086956525, 95.13, 'Age <= 53.0\\nentropy = 0.476\\nsamples = 110\\nvalue = [67, 43]'),\n",
       " Text(33.96521739130435, 67.94999999999999, 'Fare <= 27.135\\nentropy = 0.495\\nsamples = 89\\nvalue = [49, 40]'),\n",
       " Text(24.260869565217394, 40.77000000000001, 'Fare <= 26.469\\nentropy = 0.26\\nsamples = 13\\nvalue = [2, 11]'),\n",
       " Text(19.408695652173915, 13.590000000000003, 'entropy = 0.0\\nsamples = 4\\nvalue = [0, 4]'),\n",
       " Text(29.11304347826087, 13.590000000000003, 'entropy = 0.346\\nsamples = 9\\nvalue = [2, 7]'),\n",
       " Text(43.66956521739131, 40.77000000000001, 'Fare <= 387.665\\nentropy = 0.472\\nsamples = 76\\nvalue = [47, 29]'),\n",
       " Text(38.81739130434783, 13.590000000000003, 'entropy = 0.463\\nsamples = 74\\nvalue = [47, 27]'),\n",
       " Text(48.52173913043479, 13.590000000000003, 'entropy = 0.0\\nsamples = 2\\nvalue = [0, 2]'),\n",
       " Text(67.9304347826087, 67.94999999999999, 'Age <= 75.5\\nentropy = 0.245\\nsamples = 21\\nvalue = [18, 3]'),\n",
       " Text(63.07826086956522, 40.77000000000001, 'Embarked <= 0.5\\nentropy = 0.18\\nsamples = 20\\nvalue = [18, 2]'),\n",
       " Text(58.22608695652174, 13.590000000000003, 'entropy = 0.0\\nsamples = 12\\nvalue = [12, 0]'),\n",
       " Text(67.9304347826087, 13.590000000000003, 'entropy = 0.375\\nsamples = 8\\nvalue = [6, 2]'),\n",
       " Text(72.78260869565219, 40.77000000000001, 'entropy = 0.0\\nsamples = 1\\nvalue = [0, 1]'),\n",
       " Text(124.94347826086958, 122.31, 'Fare <= 51.698\\nentropy = 0.204\\nsamples = 433\\nvalue = [383, 50]'),\n",
       " Text(99.4695652173913, 95.13, 'Age <= 13.0\\nentropy = 0.193\\nsamples = 417\\nvalue = [372, 45]'),\n",
       " Text(87.33913043478262, 67.94999999999999, 'SibSp <= 2.0\\nentropy = 0.444\\nsamples = 12\\nvalue = [8, 4]'),\n",
       " Text(82.48695652173913, 40.77000000000001, 'SibSp <= 0.5\\nentropy = 0.32\\nsamples = 5\\nvalue = [1, 4]'),\n",
       " Text(77.63478260869566, 13.590000000000003, 'entropy = 0.5\\nsamples = 2\\nvalue = [1, 1]'),\n",
       " Text(87.33913043478262, 13.590000000000003, 'entropy = 0.0\\nsamples = 3\\nvalue = [0, 3]'),\n",
       " Text(92.19130434782609, 40.77000000000001, 'entropy = 0.0\\nsamples = 7\\nvalue = [7, 0]'),\n",
       " Text(111.60000000000001, 67.94999999999999, 'Age <= 32.25\\nentropy = 0.182\\nsamples = 405\\nvalue = [364, 41]'),\n",
       " Text(101.89565217391305, 40.77000000000001, 'Age <= 30.75\\nentropy = 0.206\\nsamples = 300\\nvalue = [265, 35]'),\n",
       " Text(97.04347826086958, 13.590000000000003, 'entropy = 0.185\\nsamples = 282\\nvalue = [253, 29]'),\n",
       " Text(106.74782608695654, 13.590000000000003, 'entropy = 0.444\\nsamples = 18\\nvalue = [12, 6]'),\n",
       " Text(121.30434782608697, 40.77000000000001, 'Fare <= 7.91\\nentropy = 0.108\\nsamples = 105\\nvalue = [99, 6]'),\n",
       " Text(116.45217391304348, 13.590000000000003, 'entropy = 0.0\\nsamples = 32\\nvalue = [32, 0]'),\n",
       " Text(126.15652173913044, 13.590000000000003, 'entropy = 0.151\\nsamples = 73\\nvalue = [67, 6]'),\n",
       " Text(150.41739130434783, 95.13, 'Fare <= 63.023\\nentropy = 0.43\\nsamples = 16\\nvalue = [11, 5]'),\n",
       " Text(145.56521739130437, 67.94999999999999, 'Age <= 30.0\\nentropy = 0.408\\nsamples = 7\\nvalue = [2, 5]'),\n",
       " Text(140.7130434782609, 40.77000000000001, 'Age <= 27.0\\nentropy = 0.48\\nsamples = 5\\nvalue = [2, 3]'),\n",
       " Text(135.8608695652174, 13.590000000000003, 'entropy = 0.0\\nsamples = 1\\nvalue = [0, 1]'),\n",
       " Text(145.56521739130437, 13.590000000000003, 'entropy = 0.5\\nsamples = 4\\nvalue = [2, 2]'),\n",
       " Text(150.41739130434783, 40.77000000000001, 'entropy = 0.0\\nsamples = 2\\nvalue = [0, 2]'),\n",
       " Text(155.26956521739132, 67.94999999999999, 'entropy = 0.0\\nsamples = 9\\nvalue = [9, 0]'),\n",
       " Text(232.60108695652175, 176.67000000000002, 'Pclass <= 2.5\\nentropy = 0.383\\nsamples = 314\\nvalue = [81, 233]'),\n",
       " Text(177.10434782608698, 149.49, 'Age <= 2.5\\nentropy = 0.1\\nsamples = 170\\nvalue = [9, 161]'),\n",
       " Text(164.97391304347826, 122.31, 'Pclass <= 1.5\\nentropy = 0.5\\nsamples = 2\\nvalue = [1, 1]'),\n",
       " Text(160.1217391304348, 95.13, 'entropy = 0.0\\nsamples = 1\\nvalue = [1, 0]'),\n",
       " Text(169.82608695652175, 95.13, 'entropy = 0.0\\nsamples = 1\\nvalue = [0, 1]'),\n",
       " Text(189.23478260869567, 122.31, 'Fare <= 28.856\\nentropy = 0.091\\nsamples = 168\\nvalue = [8, 160]'),\n",
       " Text(179.53043478260872, 95.13, 'Fare <= 28.231\\nentropy = 0.182\\nsamples = 69\\nvalue = [7, 62]'),\n",
       " Text(174.67826086956524, 67.94999999999999, 'Age <= 56.0\\nentropy = 0.161\\nsamples = 68\\nvalue = [6, 62]'),\n",
       " Text(164.97391304347826, 40.77000000000001, 'SibSp <= 0.5\\nentropy = 0.14\\nsamples = 66\\nvalue = [5, 61]'),\n",
       " Text(160.1217391304348, 13.590000000000003, 'entropy = 0.085\\nsamples = 45\\nvalue = [2, 43]'),\n",
       " Text(169.82608695652175, 13.590000000000003, 'entropy = 0.245\\nsamples = 21\\nvalue = [3, 18]'),\n",
       " Text(184.38260869565218, 40.77000000000001, 'Pclass <= 1.5\\nentropy = 0.5\\nsamples = 2\\nvalue = [1, 1]'),\n",
       " Text(179.53043478260872, 13.590000000000003, 'entropy = 0.0\\nsamples = 1\\nvalue = [0, 1]'),\n",
       " Text(189.23478260869567, 13.590000000000003, 'entropy = 0.0\\nsamples = 1\\nvalue = [1, 0]'),\n",
       " Text(184.38260869565218, 67.94999999999999, 'entropy = 0.0\\nsamples = 1\\nvalue = [1, 0]'),\n",
       " Text(198.9391304347826, 95.13, 'Parch <= 1.5\\nentropy = 0.02\\nsamples = 99\\nvalue = [1, 98]'),\n",
       " Text(194.08695652173915, 67.94999999999999, 'entropy = 0.0\\nsamples = 84\\nvalue = [0, 84]'),\n",
       " Text(203.7913043478261, 67.94999999999999, 'Age <= 24.5\\nentropy = 0.124\\nsamples = 15\\nvalue = [1, 14]'),\n",
       " Text(198.9391304347826, 40.77000000000001, 'entropy = 0.0\\nsamples = 10\\nvalue = [0, 10]'),\n",
       " Text(208.64347826086959, 40.77000000000001, 'Age <= 28.0\\nentropy = 0.32\\nsamples = 5\\nvalue = [1, 4]'),\n",
       " Text(203.7913043478261, 13.590000000000003, 'entropy = 0.0\\nsamples = 1\\nvalue = [1, 0]'),\n",
       " Text(213.49565217391307, 13.590000000000003, 'entropy = 0.0\\nsamples = 4\\nvalue = [0, 4]'),\n",
       " Text(288.09782608695656, 149.49, 'Fare <= 23.35\\nentropy = 0.5\\nsamples = 144\\nvalue = [72, 72]'),\n",
       " Text(270.5086956521739, 122.31, 'Age <= 36.5\\nentropy = 0.484\\nsamples = 117\\nvalue = [48, 69]'),\n",
       " Text(254.73913043478262, 95.13, 'Embarked <= 0.5\\nentropy = 0.472\\nsamples = 110\\nvalue = [42, 68]'),\n",
       " Text(237.75652173913045, 67.94999999999999, 'Fare <= 10.825\\nentropy = 0.499\\nsamples = 57\\nvalue = [27, 30]'),\n",
       " Text(228.0521739130435, 40.77000000000001, 'Fare <= 7.763\\nentropy = 0.489\\nsamples = 33\\nvalue = [19, 14]'),\n",
       " Text(223.20000000000002, 13.590000000000003, 'entropy = 0.278\\nsamples = 6\\nvalue = [1, 5]'),\n",
       " Text(232.90434782608696, 13.590000000000003, 'entropy = 0.444\\nsamples = 27\\nvalue = [18, 9]'),\n",
       " Text(247.46086956521742, 40.77000000000001, 'Fare <= 17.6\\nentropy = 0.444\\nsamples = 24\\nvalue = [8, 16]'),\n",
       " Text(242.60869565217394, 13.590000000000003, 'entropy = 0.231\\nsamples = 15\\nvalue = [2, 13]'),\n",
       " Text(252.31304347826088, 13.590000000000003, 'entropy = 0.444\\nsamples = 9\\nvalue = [6, 3]'),\n",
       " Text(271.7217391304348, 67.94999999999999, 'Age <= 29.75\\nentropy = 0.406\\nsamples = 53\\nvalue = [15, 38]'),\n",
       " Text(266.8695652173913, 40.77000000000001, 'Fare <= 15.373\\nentropy = 0.38\\nsamples = 51\\nvalue = [13, 38]'),\n",
       " Text(262.0173913043478, 13.590000000000003, 'entropy = 0.439\\nsamples = 40\\nvalue = [13, 27]'),\n",
       " Text(271.7217391304348, 13.590000000000003, 'entropy = 0.0\\nsamples = 11\\nvalue = [0, 11]'),\n",
       " Text(276.5739130434783, 40.77000000000001, 'entropy = 0.0\\nsamples = 2\\nvalue = [2, 0]'),\n",
       " Text(286.27826086956526, 95.13, 'Age <= 55.0\\nentropy = 0.245\\nsamples = 7\\nvalue = [6, 1]'),\n",
       " Text(281.4260869565218, 67.94999999999999, 'entropy = 0.0\\nsamples = 6\\nvalue = [6, 0]'),\n",
       " Text(291.13043478260875, 67.94999999999999, 'entropy = 0.0\\nsamples = 1\\nvalue = [0, 1]'),\n",
       " Text(305.68695652173915, 122.31, 'Parch <= 0.5\\nentropy = 0.198\\nsamples = 27\\nvalue = [24, 3]'),\n",
       " Text(300.83478260869566, 95.13, 'entropy = 0.0\\nsamples = 1\\nvalue = [0, 1]'),\n",
       " Text(310.53913043478263, 95.13, 'Age <= 5.5\\nentropy = 0.142\\nsamples = 26\\nvalue = [24, 2]'),\n",
       " Text(300.83478260869566, 67.94999999999999, 'Fare <= 31.331\\nentropy = 0.444\\nsamples = 3\\nvalue = [2, 1]'),\n",
       " Text(295.9826086956522, 40.77000000000001, 'entropy = 0.0\\nsamples = 2\\nvalue = [2, 0]'),\n",
       " Text(305.68695652173915, 40.77000000000001, 'entropy = 0.0\\nsamples = 1\\nvalue = [0, 1]'),\n",
       " Text(320.2434782608696, 67.94999999999999, 'Parch <= 4.5\\nentropy = 0.083\\nsamples = 23\\nvalue = [22, 1]'),\n",
       " Text(315.3913043478261, 40.77000000000001, 'entropy = 0.0\\nsamples = 18\\nvalue = [18, 0]'),\n",
       " Text(325.0956521739131, 40.77000000000001, 'Age <= 38.5\\nentropy = 0.32\\nsamples = 5\\nvalue = [4, 1]'),\n",
       " Text(320.2434782608696, 13.590000000000003, 'entropy = 0.0\\nsamples = 1\\nvalue = [0, 1]'),\n",
       " Text(329.9478260869565, 13.590000000000003, 'entropy = 0.0\\nsamples = 4\\nvalue = [4, 0]')]"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#Separando features para segundo treinamento com profundidade máxima na árvore\n",
    "feat_dois = treinamento[nomes_feat].values\n",
    "arvore_decisao_dois = tree.DecisionTreeClassifier(\n",
    "    max_depth = 7,\n",
    "    random_state = 1\n",
    ")\n",
    "\n",
    "arvore_decisao_dois = arvore_decisao_dois.fit(feat_dois, alvo)\n",
    "print(arvore_decisao_dois.feature_importances_)\n",
    "print(arvore_decisao_dois.score(feat_dois, alvo))\n",
    "tree.plot_tree(arvore_decisao_dois, feature_names=nomes_feat)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.625      0.78688525 0.61016949 0.83333333 0.82539683 0.72131148\n",
      " 0.72413793 0.62068966 0.8        0.82352941]\n",
      "0.7370453369538661\n"
     ]
    }
   ],
   "source": [
    "#validação cruzada para checar a precisão do modelo\n",
    "scores = model_selection.cross_val_score(arvore_decisao_dois,feat_dois, alvo, scoring='f1', cv=10)\n",
    "print(scores)\n",
    "print(scores.mean())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      " Nova previsão\n"
     ]
    }
   ],
   "source": [
    "print(\"\\n Nova previsão\")\n",
    "\n",
    "teste_feat_dois = teste[[\"Pclass\", \"Age\", \"Fare\", \"Embarked\", \"Sex\", \"SibSp\", \"Parch\"]].values\n",
    "#realizando nova precisão baseada na segunda árvore de decisões e salvando e .csv\n",
    "previsao_dois = arvore_decisao_dois.predict(teste_feat_dois)\n",
    "limparDados.escrever_previsao(previsao_dois, \"arvore_decisao_dois.csv\")"
   ]
  }
 ],
 "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}