exemplos/basic_use_machine_learning.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# Machine Learning\n",
    "\n",
    "<!-- executar como ppt: jupyter-nbconvert ElipseML\\init_ppt.ipynb --to slides --post serve   -->"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "## O que devemos saber\n",
    "- Existe a bastante tempo mas só recemente vem se popularizando\n",
    "- Companhias querem usar ML, mas não possuem certeza sobre o que é isso ou aonde aplicar\n",
    "- Possui um grande potencial mas não da resultados instantaneos, é preciso responder as perguntas certas e planejar cuidadosamente o projeto\n",
    "- Tenha certeza que você possui: dados certos, pessoas e processos. Se tudo estiver certo, tenha certeza que um projeto de ML pode salvar boas quantias de dinheiro e aumetar a produtividade após algum meses de desenvolvimento!\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "## Aplicações\n",
    "### Analises e Manutenção Preditivas\n",
    "- Falhas e Configuração de Equipamentos\n",
    "- Controle de Qualidade\n",
    "- Quais ações são mais efetivas\n",
    "- Escolha de investimentos e parceiros\n",
    "- Previsão de demanda\n",
    "- Treinamentos de robôs"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "## Passos para Aplicação de ML\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Dados\n",
    "- Coletar fontes diferentes. Históricos, ERP, etc.\n",
    "- Coleta automizada\n",
    "- Qualidade é mais importante que quantidade"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Classificar Dados\n",
    "- Pessoas domínio do processo para classificar os dados como bons ou ruins"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Escolher uma questão\n",
    "- Começar com o que você realmente quer saber\n",
    "- Função de custo"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Usar conhecimento do processo (Domain Knowledge)\n",
    "- Quais dados podem responder sua questão?\n",
    "- É possível adquirir dados nulos?\n",
    "- Qual a qualidade dos dados?\n",
    "- Eliminar variáveis dependentes. Mais atributos = Modelo mais complexo.\n",
    "\n",
    "<input type='hidden' value='https://mineracaodedados.wordpress.com/2015/06/13/7-tecnicas-para-reducao-da-dimensionalidade/'>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### ETL (Extract, Transform, Load)\n",
    "- Posso automatizar cada passo?\n",
    "- Posso obter novos dados automaticamente?\n",
    "- O quão limpos os dados devem estar?\n",
    "- Como os valores nulos serão manipulados? Remoção completa, substituído pela média ou pelo valor recorrente?"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Visualizar seus dados\n",
    "Procurar por:\n",
    "-  Problemas nos dados\n",
    "-  Tendências\n",
    "-  Algoritmo"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Qual algoritmo usar?\n",
    "- Classificação ou Regressão?\n",
    "- Dados Classificados ou não classificados?\n",
    "- Regras de associação\n",
    "- Tolerante a dados nulos?\n",
    "- Aprendizado Incremental/Tardio (Lazy Learners)? Gerar um novo modelo(retraining)?\n",
    "- Black box vs Maior legibilidade?\n",
    "- Tolerante a outliers?"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Qual plataforma utilizar?\n",
    "- Computação na nuvem ou local?\n",
    "- Quão flexível é?\n",
    "- Quão fácil é obter dados nessa plataforma?\n",
    "- Quão fácil é obter resultados utilizáveis? Poder ser automatizado?"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Testar o modelo\n",
    "- Técnicas de testes\n",
    "- Quão preciso é?\n",
    "- Possuir históricos dos passos e reavaliar quando necessário"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# Como Funciona\n",
    "\n",
    "## Treinamento e Criação do Modelo!\n",
    "\n",
    "<img src=\"images/fig_ml_01.png\" />"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "## Predição pelo Computador\n",
    "\n",
    "<img src=\"images/fig_ml_02.png\" />"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# Machine Learning em Python\n",
    "\n",
    "## Exemplo utilizando a biblioteca [scikit-learn](http://scikit-learn.org/stable/) e algoritmo [KNN](http://scikit-learn.org/stable/modules/neighbors.html)\n",
    "\n",
    "### Dataset utilizado: [Boston house prices](http://scikit-learn.org/stable/datasets/index.html#boston-house-prices-dataset)\n",
    "\n",
    "Objetivo do exemplo é predizer o valor de um imóvel em Boston utilizando dados já existentes referente a outros imóveis."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "outputs": [],
   "source": [
    "from sklearn.datasets import load_boston\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.neighbors import KNeighborsRegressor\n",
    "\n",
    "boston = load_boston()\n",
    "#numero de vizinhos próximos\n",
    "K = 9\n",
    "#algoritmo para ser utilizado\n",
    "knn = KNeighborsRegressor(n_neighbors=K)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Separa os dados para serem utilizado no treinamento do algoritimo\n",
    "\n",
    "* x são os dados de entrada\n",
    "* y valor do imóvel"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "#numero de registros utilizado para gerar o modelo\n",
    "n_data = 100\n",
    "\n",
    "x, y = boston.data[:n_data], boston.target[:n_data]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Executa o Treino com os dados"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "KNeighborsRegressor(algorithm='auto', leaf_size=30, metric='minkowski',\n",
       "          metric_params=None, n_jobs=None, n_neighbors=9, p=2,\n",
       "          weights='uniform')"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "knn.fit(x, y)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Executa a Predição"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "valor_predicao = knn.predict([boston.data[0]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Valor Real:  24.0\n",
      "Valor Estimado:  21.08888888888889\n"
     ]
    }
   ],
   "source": [
    "valor_real= boston.target[0]\n",
    "\n",
    "print(\"Valor Real: \",valor_real)\n",
    "print(\"Valor Estimado: \",valor_predicao[0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Comparação entre Valores Preditos x Valores Reais\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "outputs": [
    {
     "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": [
    "import numpy as np\n",
    "\n",
    "#Predizendo N_data valores\n",
    "y_predicted = knn.predict(x)\n",
    "\n",
    "#Visualização gráfica\n",
    "plt.plot(np.linspace(-1, 1, n_data), y, label='Valores Reais', color='black')\n",
    "plt.plot(np.linspace(-1, 1, n_data), y_predicted, label='Valores Preditos', color='red')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Ajustando parâmetros do algoritimo, como por exemplo o K, ou utilizando mais dados a precisão do modelo pode aumentar ou diminuir."
   ]
  }
 ],
 "metadata": {
  "celltoolbar": "Slideshow",
  "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": "3.6.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}