{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 一.原理介绍\n",
    "这一节将树模型的预测与概率分布相结合，我们假设树模型的输出服从某一分布，而我们的目标是使得该输出的概率尽可能的高，如下图所示\n",
    "![avatar](./source/10_集成学习_极大似然估计.png)\n",
    "\n",
    "而概率值最高的点通常由分布中的某一个参数（通常是均值）反映，所以我们将树模型的输出打造为分布中的该参数项，然后让树模型的输出去逼近极大似然估计的结果即可，即：   \n",
    "\n",
    "$$\n",
    "\\hat{y}\\rightarrow \\mu_{ML}\n",
    "$$   \n",
    "\n",
    "下面分别介绍possion回归，gamma回归，tweedie回归，负二项回归的具体求解"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 二.泊松回归\n",
    "泊松分布的表达式如下：    \n",
    "\n",
    "$$\n",
    "P(y\\mid\\lambda)=\\frac{\\lambda^y}{y!}e^{-\\lambda}\n",
    "$$  \n",
    "\n",
    "其中，$y$是我们的目标输出，$\\lambda$为模型参数，且$\\lambda$恰为该分布的均值，由于泊松分布要求$y>0$，所以我们对$\\hat{y}$取指数去拟合$\\lambda$，即令：   \n",
    "\n",
    "$$\n",
    "\\lambda=e^{\\hat{y}}\n",
    "$$   \n",
    "\n",
    "对于$N$个样本，其似然函数可以表示如下：  \n",
    "\n",
    "$$\n",
    "\\prod_{i=1}^N\\frac{e^{y_i\\hat{y_i}}e^{-e^{\\hat{y_i}}}}{y_i!}\n",
    "$$   \n",
    "\n",
    "由于$y_i!$是常数，可以去掉，并对上式取负对数，转换为求极小值的问题：  \n",
    "\n",
    "$$\n",
    "L(y,\\hat{y})=\\sum_{i=1}^N(e^{\\hat{y_i}}-y_i\\hat{y_i})\n",
    "$$  \n",
    "\n",
    "所以，一阶导和二阶导分别为：   \n",
    "\n",
    "$$\n",
    "\\frac{\\partial L(y,\\hat{y})}{\\partial \\hat{y}}=e^{\\hat{y}}-y\\\\\n",
    "\\frac{\\partial^2 L(y,\\hat{y})}{{\\partial \\hat{y}}^2}=e^{\\hat{y}}\\\\\n",
    "$$   "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 三.gamma回归\n",
    "gamma分布如下：   \n",
    "\n",
    "$$\n",
    "p(y\\mid\\alpha,\\lambda)=\\frac{1}{\\Gamma(\\alpha)\\lambda^\\alpha}y^{\\alpha-1}e^{-y/\\lambda}\n",
    "$$  \n",
    "\n",
    "其中，$y>0$为我们的目标输出，$\\alpha$为形状参数，$\\lambda$为尺度参数，$\\Gamma(\\cdot)$为Gamma函数（后续推导这里会被省略，所以就不列出来了），而Gamma分布的均值为$\\alpha\\lambda$，这里不好直接变换，我们令$\\alpha=1/\\phi,\\lambda=\\phi\\mu$，所以现在Gamma分布的均值可以表示为$\\mu$，此时的Gamma分布为：   \n",
    "\n",
    "$$\n",
    "p(y\\mid\\mu,\\phi)=\\frac{1}{y\\Gamma(1/\\phi)}(\\frac{y}{\\mu\\phi})^{1/\\phi}exp[-\\frac{y}{\\mu\\phi}]\n",
    "$$  \n",
    "\n",
    "此时，$\\mu$看做Gamma分布的均值参数，而$\\phi$为它的离散参数，在均值给定的情况下，若离散参数越大，Gamma分布的离散程度越大，接下来对上面的表达式进一步变换：   \n",
    "\n",
    "$$\n",
    "p(y\\mid\\mu,\\phi)=exp[\\frac{-y/\\mu-ln\\mu}{\\phi}+\\frac{1-\\phi}{\\phi}lny-\\frac{ln\\phi}{\\phi}-ln\\Gamma(\\frac{1}{\\phi})]\n",
    "$$  \n",
    "\n",
    "同泊松分布一样，我们可以令：   \n",
    "\n",
    "$$\n",
    "\\mu=e^{\\hat{y}}\n",
    "$$  \n",
    "\n",
    "又由于$\\mu$与$\\phi$无关，所以做极大似然估计时可以将$\\phi$看做常数，我们将对数似然函数的负数看做损失函数，可以写作如下：   \n",
    "\n",
    "$$\n",
    "L(y,\\hat{y})=\\sum_{i=1}^N(\\frac{y_i}{e^{\\hat{y_i}}}+\\hat{y_i})\n",
    "$$  \n",
    "\n",
    "所以，一阶导和二阶导就可以写出来啦：   \n",
    "\n",
    "$$\n",
    "\\frac{\\partial L(y,\\hat{y})}{\\partial \\hat{y}}=1-ye^{-\\hat{y}}\\\\\n",
    "\\frac{\\partial^2 L(y,\\hat{y})}{{\\partial \\hat{y}}^2}=ye^{-\\hat{y}}\\\\\n",
    "$$  \n",
    "\n",
    "注意：上面的两个向量是按元素相乘"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 四.tweedie回归\n",
    "tweedie回归是多个分布的组合体，包括gamma分布，泊松分布，高斯分布等，tweedie回归由一个超参数$p$控制，$p$不同，则其对应的对数似然函数也不同：   \n",
    "\n",
    "$$\n",
    "g(y,\\phi)+\\left\\{\\begin{matrix}\n",
    "\\frac{1}{\\phi}(ylog(\\mu)-\\frac{\\mu^{2-p}}{2-p}) &  p=1\\\\ \n",
    "\\frac{1}{\\phi}(y\\frac{\\mu^{1-p}}{1-p}-log\\mu) & p=2 \\\\ \n",
    "\\frac{1}{\\phi}(y\\frac{\\mu^{1-p}}{1-p}-\\frac{\\mu^{2-p}}{2-p}) & p\\neq 1,p\\neq 2\n",
    "\\end{matrix}\\right.\n",
    "$$  \n",
    "\n",
    "同样的，我们可以令：   \n",
    "\n",
    "$$\n",
    "\\mu=e^{\\hat{y}}\n",
    "$$  \n",
    "\n",
    "由于除开$\\mu$以外的都可以视作常数项，所以损失函数可以简化为：   \n",
    "\n",
    "$$\n",
    "L(y,\\hat{y})=\\left\\{\\begin{matrix}\n",
    "\\sum_{i=1}^n(\\frac{e^{\\hat{y_i}(2-p)}}{2-p}-y_i\\hat{y_i})=\\sum_{i=1}^n(e^{\\hat{y_i}}-y_i\\hat{y_i}) &  p=1\\\\ \n",
    "\\sum_{i=1}^n(\\hat{y_i}+y_ie^{-\\hat{y_i}}) & p=2 \\\\ \n",
    "\\sum_{i=1}^n(\\frac{exp[\\hat{y_i}(2-p)]}{2-p}-y_i\\frac{exp[\\hat{y_i}(1-p)]}{1-p}) & p\\neq 1,p\\neq 2\n",
    "\\end{matrix}\\right.\n",
    "$$   \n",
    "\n",
    "所以，一阶导：   \n",
    "\n",
    "$$\n",
    "\\frac{\\partial L(y,\\hat{y})}{\\partial \\hat{y}}=\\left\\{\\begin{matrix}\n",
    "e^{\\hat{y}}-y &  p=1\\\\ \n",
    "1-ye^{-\\hat{y}} & p=2 \\\\ \n",
    "e^{\\hat{y}(2-p)}-ye^{\\hat{y}(1-p)} & p\\neq 1,p\\neq 2\n",
    "\\end{matrix}\\right.\n",
    "$$\n",
    "\n",
    "二阶导：   \n",
    "\n",
    "$$\n",
    "\\frac{\\partial^2 L(y,\\hat{y})}{{\\partial \\hat{y}}^2}=\\left\\{\\begin{matrix}\n",
    "e^{\\hat{y}} &  p=1\\\\ \n",
    "ye^{-\\hat{y}} & p=2 \\\\ \n",
    "(2-p)e^{\\hat{y}(2-p)}-(1-p)ye^{\\hat{y}(1-p)} & p\\neq 1,p\\neq 2\n",
    "\\end{matrix}\\right.\n",
    "$$"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 五.代码实现\n",
    "基于上一节的计算框架，略作调整即可实现...."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "os.chdir('../')\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "from ml_models.ensemble import XGBoostBaseTree\n",
    "from ml_models import utils\n",
    "import copy\n",
    "import numpy as np\n",
    "\n",
    "\"\"\"\n",
    "xgboost回归树的实现，封装到ml_models.ensemble\n",
    "\"\"\"\n",
    "\n",
    "class XGBoostRegressor(object):\n",
    "    def __init__(self, base_estimator=None, n_estimators=10, learning_rate=1.0, loss='squarederror', p=2.5):\n",
    "        \"\"\"\n",
    "        :param base_estimator: 基学习器\n",
    "        :param n_estimators: 基学习器迭代数量\n",
    "        :param learning_rate: 学习率，降低后续基学习器的权重，避免过拟合\n",
    "        :param loss:损失函数，支持squarederror、logistic、poisson,gamma,tweedie\n",
    "        :param p:对tweedie回归生效\n",
    "        \"\"\"\n",
    "        self.base_estimator = base_estimator\n",
    "        self.n_estimators = n_estimators\n",
    "        self.learning_rate = learning_rate\n",
    "        if self.base_estimator is None:\n",
    "            # 默认使用决策树桩\n",
    "            self.base_estimator = XGBoostBaseTree()\n",
    "        # 同质分类器\n",
    "        if type(base_estimator) != list:\n",
    "            estimator = self.base_estimator\n",
    "            self.base_estimator = [copy.deepcopy(estimator) for _ in range(0, self.n_estimators)]\n",
    "        # 异质分类器\n",
    "        else:\n",
    "            self.n_estimators = len(self.base_estimator)\n",
    "        self.loss = loss\n",
    "        self.p = p\n",
    "\n",
    "    def _get_gradient_hess(self, y, y_pred):\n",
    "        \"\"\"\n",
    "        获取一阶、二阶导数信息\n",
    "        :param y:真实值\n",
    "        :param y_pred:预测值\n",
    "        :return:\n",
    "        \"\"\"\n",
    "        if self.loss == 'squarederror':\n",
    "            return y_pred - y, np.ones_like(y)\n",
    "        elif self.loss == 'logistic':\n",
    "            return utils.sigmoid(y_pred) - utils.sigmoid(y), utils.sigmoid(y_pred) * (1 - utils.sigmoid(y_pred))\n",
    "        elif self.loss == 'poisson':\n",
    "            return np.exp(y_pred) - y, np.exp(y_pred)\n",
    "        elif self.loss == 'gamma':\n",
    "            return 1.0 - y * np.exp(-1.0 * y_pred), y * np.exp(-1.0 * y_pred)\n",
    "        elif self.loss == 'tweedie':\n",
    "            if self.p == 1:\n",
    "                return np.exp(y_pred) - y, np.exp(y_pred)\n",
    "            elif self.p == 2:\n",
    "                return 1.0 - y * np.exp(-1.0 * y_pred), y * np.exp(-1.0 * y_pred)\n",
    "            else:\n",
    "                return np.exp(y_pred * (2.0 - self.p)) - y * np.exp(y_pred * (1.0 - self.p)), (2.0 - self.p) * np.exp(\n",
    "                    y_pred * (2.0 - self.p)) - (1.0 - self.p) * y * np.exp(y_pred * (1.0 - self.p))\n",
    "\n",
    "    def fit(self, x, y):\n",
    "        y_pred = np.zeros_like(y)\n",
    "        g, h = self._get_gradient_hess(y, y_pred)\n",
    "        for index in range(0, self.n_estimators):\n",
    "            self.base_estimator[index].fit(x, g, h)\n",
    "            y_pred += self.base_estimator[index].predict(x) * self.learning_rate\n",
    "            g, h = self._get_gradient_hess(y, y_pred)\n",
    "\n",
    "    def predict(self, x):\n",
    "        rst_np = np.sum(\n",
    "            [self.base_estimator[0].predict(x)] +\n",
    "            [self.learning_rate * self.base_estimator[i].predict(x) for i in\n",
    "             range(1, self.n_estimators - 1)] +\n",
    "            [self.base_estimator[self.n_estimators - 1].predict(x)]\n",
    "            , axis=0)\n",
    "        if self.loss in [\"poisson\", \"gamma\", \"tweedie\"]:\n",
    "            return np.exp(rst_np)\n",
    "        else:\n",
    "            return rst_np"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "对泊松、gamma、tweedie回归做测试"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = np.linspace(1, 10, num=100)\n",
    "target = np.sin(data) + np.random.random(size=100) + 1  # 添加噪声\n",
    "data = data.reshape((-1, 1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x29c33c1a9b0>]"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x29c33c43e80>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "model = XGBoostRegressor(loss='poisson')\n",
    "model.fit(data, target)\n",
    "plt.scatter(data, target)\n",
    "plt.plot(data, model.predict(data), color='r')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x29c33c1a400>]"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x29c33bfbe80>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "model = XGBoostRegressor(loss='gamma')\n",
    "model.fit(data, target)\n",
    "plt.scatter(data, target)\n",
    "plt.plot(data, model.predict(data), color='r')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x29c33c1aba8>]"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x29c33c36be0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "model = XGBoostRegressor(loss='tweedie',p=2.5)\n",
    "model.fit(data, target)\n",
    "plt.scatter(data, target)\n",
    "plt.plot(data, model.predict(data), color='r')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x29c33c43d68>]"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x29c33b6f128>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "model = XGBoostRegressor(loss='tweedie',p=1.5)\n",
    "model.fit(data, target)\n",
    "plt.scatter(data, target)\n",
    "plt.plot(data, model.predict(data), color='r')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "上面的拟合结果，看不出明显区别....,接下来对tweedie分布中`p`取极端值做一个简单探索...，可以发现取值过大或者过小都有可能陷入欠拟合"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x29c44ed2cc0>]"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x29c44daf208>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "model = XGBoostRegressor(loss='tweedie',p=0.1)\n",
    "model.fit(data, target)\n",
    "plt.scatter(data, target)\n",
    "plt.plot(data, model.predict(data), color='r')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x29c44e94f28>]"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x29c44e41390>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "model = XGBoostRegressor(loss='tweedie',p=20)\n",
    "model.fit(data, target)\n",
    "plt.scatter(data, target)\n",
    "plt.plot(data, model.predict(data), color='r')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "参考内容：  \n",
    "http://www.doc88.com/p-9029670237688.html  \n",
    "http://docs.h2o.ai/h2o/latest-stable/h2o-docs/data-science/glm.html#families"
   ]
  }
 ],
 "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": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}