{ "cells": [ { "cell_type": "markdown", "metadata": {}, "source": [ "# 机器学习100天——第3天：多元线性回归（Multiple Linear Regression）" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 第1步：数据预处理" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "**导入库**" ] }, { "cell_type": "code", "execution_count": 1, "metadata": {}, "outputs": [], "source": [ "import pandas as pd\n", "import numpy as np" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "**导入数据集**" ] }, { "cell_type": "code", "execution_count": 30, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "[[165349.2 136897.8 471784.1 'New York']\n", " [162597.7 151377.59 443898.53 'California']\n", " [153441.51 101145.55 407934.54 'Florida']\n", " [144372.41 118671.85 383199.62 'New York']\n", " [142107.34 91391.77 366168.42 'Florida']\n", " [131876.9 99814.71 362861.36 'New York']\n", " [134615.46 147198.87 127716.82 'California']\n", " [130298.13 145530.06 323876.68 'Florida']\n", " [120542.52 148718.95 311613.29 'New York']\n", " [123334.88 108679.17 304981.62 'California']]\n", "[192261.83 191792.06 191050.39 182901.99 166187.94 156991.12 156122.51\n", " 155752.6 152211.77 149759.96 146121.95 144259.4 141585.52 134307.35\n", " 132602.65 129917.04 126992.93 125370.37 124266.9 122776.86 118474.03\n", " 111313.02 110352.25 108733.99 108552.04 107404.34 105733.54 105008.31\n", " 103282.38 101004.64 99937.59 97483.56 97427.84 96778.92 96712.8\n", " 96479.51 90708.19 89949.14 81229.06 81005.76 78239.91 77798.83\n", " 71498.49 69758.98 65200.33 64926.08 49490.75 42559.73 35673.41\n", " 14681.4 ]\n" ] } ], "source": [ "dataset = pd.read_csv('../datasets/50_Startups.csv')\n", "X = dataset.iloc[ : , :-1].values\n", "Y = dataset.iloc[ : , 4 ].values\n", "print(X[:10])\n", "print(Y)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "**将类别数据数字化**" ] }, { "cell_type": "code", "execution_count": 31, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "labelencoder:\n", "[[165349.2 136897.8 471784.1 2]\n", " [162597.7 151377.59 443898.53 0]\n", " [153441.51 101145.55 407934.54 1]\n", " [144372.41 118671.85 383199.62 2]\n", " [142107.34 91391.77 366168.42 1]\n", " [131876.9 99814.71 362861.36 2]\n", " [134615.46 147198.87 127716.82 0]\n", " [130298.13 145530.06 323876.68 1]\n", " [120542.52 148718.95 311613.29 2]\n", " [123334.88 108679.17 304981.62 0]]\n", "onehot:\n", "[[0.0000000e+00 0.0000000e+00 1.0000000e+00 1.6534920e+05 1.3689780e+05\n", " 4.7178410e+05]\n", " [1.0000000e+00 0.0000000e+00 0.0000000e+00 1.6259770e+05 1.5137759e+05\n", " 4.4389853e+05]\n", " [0.0000000e+00 1.0000000e+00 0.0000000e+00 1.5344151e+05 1.0114555e+05\n", " 4.0793454e+05]\n", " [0.0000000e+00 0.0000000e+00 1.0000000e+00 1.4437241e+05 1.1867185e+05\n", " 3.8319962e+05]\n", " [0.0000000e+00 1.0000000e+00 0.0000000e+00 1.4210734e+05 9.1391770e+04\n", " 3.6616842e+05]\n", " [0.0000000e+00 0.0000000e+00 1.0000000e+00 1.3187690e+05 9.9814710e+04\n", " 3.6286136e+05]\n", " [1.0000000e+00 0.0000000e+00 0.0000000e+00 1.3461546e+05 1.4719887e+05\n", " 1.2771682e+05]\n", " [0.0000000e+00 1.0000000e+00 0.0000000e+00 1.3029813e+05 1.4553006e+05\n", " 3.2387668e+05]\n", " [0.0000000e+00 0.0000000e+00 1.0000000e+00 1.2054252e+05 1.4871895e+05\n", " 3.1161329e+05]\n", " [1.0000000e+00 0.0000000e+00 0.0000000e+00 1.2333488e+05 1.0867917e+05\n", " 3.0498162e+05]]\n" ] } ], "source": [ "from sklearn.preprocessing import LabelEncoder, OneHotEncoder\n", "labelencoder = LabelEncoder()\n", "X[: , 3] = labelencoder.fit_transform(X[ : , 3])\n", "print(\"labelencoder:\")\n", "print(X[:10])\n", "onehotencoder = OneHotEncoder(categorical_features = [3])\n", "X = onehotencoder.fit_transform(X).toarray()\n", "print(\"onehot:\")\n", "print(X[:10])" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "**躲避虚拟变量陷阱**\n", "\n", "在回归预测中我们需要所有的数据都是numeric的，但是会有一些非numeric的数据，比如国家，省，部门，性别。这时候我们需要设置虚拟变量（Dummy variable）。做法是将此变量中的每一个值，衍生成为新的变量，是设为1，否设为0.举个例子，“性别”这个变量,我们可以虚拟出“男”和”女”两虚拟变量，男性的话“男”值为1，”女”值为0；女性的话“男”值为0，”女”值为1。\n", "\n", "但是要注意，这时候虚拟变量陷阱就出现了。就拿性别来说，其实一个虚拟变量就够了，比如 1 的时候是“男”， 0 的时候是”非男”，即为女。如果设置两个虚拟变量“男”和“女”，语义上来说没有问题，可以理解，但是在回归预测中会多出一个变量，多出的这个变量将会对回归预测结果产生影响。一般来说，如果虚拟变量要比实际变量的种类少一个。 \n", "\n", "在多重线性回归中，变量不是越多越好，而是选择适合的变量。这样才会对结果准确预测。如果category类的特征都放进去，拟合的时候，所有权重的计算，都可以有两种方法实现，一种是提高某个category的w，一种是降低其他category的w，这两种效果是等效的，也就是发生了共线性,虚拟变量系数相加和为1，出现完全共线陷阱。\n", "\n", "**但是下面测试尽然和想法不一致。。。**" ] }, { "cell_type": "code", "execution_count": 32, "metadata": {}, "outputs": [], "source": [ "X1 = X[: , 1:]" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "**拆分数据集为训练集和测试集**" ] }, { "cell_type": "code", "execution_count": 39, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "[[0.0000000e+00 1.0000000e+00 0.0000000e+00 6.6051520e+04 1.8264556e+05\n", " 1.1814820e+05]\n", " [1.0000000e+00 0.0000000e+00 0.0000000e+00 1.0067196e+05 9.1790610e+04\n", " 2.4974455e+05]\n", " [0.0000000e+00 1.0000000e+00 0.0000000e+00 1.0191308e+05 1.1059411e+05\n", " 2.2916095e+05]\n", " [0.0000000e+00 1.0000000e+00 0.0000000e+00 2.7892920e+04 8.4710770e+04\n", " 1.6447071e+05]\n", " [0.0000000e+00 1.0000000e+00 0.0000000e+00 1.5344151e+05 1.0114555e+05\n", " 4.0793454e+05]\n", " [0.0000000e+00 0.0000000e+00 1.0000000e+00 7.2107600e+04 1.2786455e+05\n", " 3.5318381e+05]\n", " [0.0000000e+00 0.0000000e+00 1.0000000e+00 2.0229590e+04 6.5947930e+04\n", " 1.8526510e+05]\n", " [0.0000000e+00 0.0000000e+00 1.0000000e+00 6.1136380e+04 1.5270192e+05\n", " 8.8218230e+04]\n", " [0.0000000e+00 1.0000000e+00 0.0000000e+00 7.3994560e+04 1.2278275e+05\n", " 3.0331926e+05]\n", " [0.0000000e+00 1.0000000e+00 0.0000000e+00 1.4210734e+05 9.1391770e+04\n", " 3.6616842e+05]]\n", "[103282.38 144259.4 146121.95 77798.83 191050.39 105008.31 81229.06\n", " 97483.56 110352.25 166187.94]\n", "[[1.0000000e+00 0.0000000e+00 6.6051520e+04 1.8264556e+05 1.1814820e+05]\n", " [0.0000000e+00 0.0000000e+00 1.0067196e+05 9.1790610e+04 2.4974455e+05]\n", " [1.0000000e+00 0.0000000e+00 1.0191308e+05 1.1059411e+05 2.2916095e+05]\n", " [1.0000000e+00 0.0000000e+00 2.7892920e+04 8.4710770e+04 1.6447071e+05]\n", " [1.0000000e+00 0.0000000e+00 1.5344151e+05 1.0114555e+05 4.0793454e+05]\n", " [0.0000000e+00 1.0000000e+00 7.2107600e+04 1.2786455e+05 3.5318381e+05]\n", " [0.0000000e+00 1.0000000e+00 2.0229590e+04 6.5947930e+04 1.8526510e+05]\n", " [0.0000000e+00 1.0000000e+00 6.1136380e+04 1.5270192e+05 8.8218230e+04]\n", " [1.0000000e+00 0.0000000e+00 7.3994560e+04 1.2278275e+05 3.0331926e+05]\n", " [1.0000000e+00 0.0000000e+00 1.4210734e+05 9.1391770e+04 3.6616842e+05]]\n", "[103282.38 144259.4 146121.95 77798.83 191050.39 105008.31 81229.06\n", " 97483.56 110352.25 166187.94]\n" ] } ], "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 = 0.2, random_state = 0)\n", "X1_train, X1_test, Y1_train, Y1_test = train_test_split(X1, Y, test_size = 0.2, random_state = 0)\n", "print(X_test)\n", "print(Y_test)\n", "print(X1_test)\n", "print(Y1_test)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 第2步：在训练集上训练多元线性回归模型" ] }, { "cell_type": "code", "execution_count": 40, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "LinearRegression(copy_X=True, fit_intercept=True, n_jobs=1, normalize=False)" ] }, "execution_count": 40, "metadata": {}, "output_type": "execute_result" } ], "source": [ "from sklearn.linear_model import LinearRegression\n", "regressor = LinearRegression()\n", "regressor.fit(X_train, Y_train)\n", "regressor1 = LinearRegression()\n", "regressor1.fit(X1_train, Y1_train)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 第3步：在测试集上预测结果¶" ] }, { "cell_type": "code", "execution_count": 41, "metadata": {}, "outputs": [], "source": [ "y_pred = regressor.predict(X_test)\n", "y1_pred = regressor1.predict(X1_test)" ] }, { "cell_type": "code", "execution_count": 42, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "[103015.20159796 132582.27760815 132447.73845173 71976.09851258\n", " 178537.48221051 116161.24230163 67851.69209676 98791.73374689\n", " 113969.43533011 167921.06569547]\n", "[103015.20159795 132582.27760817 132447.73845176 71976.09851257\n", " 178537.48221058 116161.24230165 67851.69209675 98791.73374686\n", " 113969.43533013 167921.06569553]\n" ] } ], "source": [ "print(y_pred)\n", "print(y1_pred)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "**完整的项目请前往Github项目100-Days-Of-ML-Code查看。有任何的建议或者意见欢迎在issue中提出~**" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [] } ], "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": "3.6.5" } }, "nbformat": 4, "nbformat_minor": 2 }