SOLUTION2.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 任务2 - 特征工程（2天）\n",
    "\n",
    "任务2：对数据特征进行衍生和进行特征挑选。时间：2天\n",
    "\n",
    "  * 特征衍生\n",
    "\n",
    "  * 特征挑选：分别用IV值和随机森林等进行特征选择\n",
    "\n",
    "  * ……以及你能想到特征工程处理"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 特征衍生\n",
    "  在实际业务中，通常我们只拥有几个到几十个不等的基础变量，而多数变量没有实际含义，不适合直接建模，如用户地址（多种属性值的分类变量）、用户日消费金额（弱数值变量）。而此类变量在做一定的变换或者组合后，往往具有较强的信息价值，对数据敏感性和机器学习实战经验能起到一定的帮助作用。所以我们需要对基础特征做一些衍生类的工作，也就是业内常说的如何生成万维数据。\n",
    "\n",
    "  特征衍生也叫特征构建，是指从原始数据中构建新的特征，也属于特征选择的一种手段。特征构建工作并不完全依赖于技术，它要求我们具备相关领域丰富的知识或者实践经验，基于业务，花时间去观察和分析原始数据，思考问题的潜在形式和数据结构，从原始数据中找出一些具有物理意义的特征。\n",
    "\n",
    "  找到可以拓展的基础特征后，便可用如下几种方式衍生特征：\n",
    "\n",
    "特征扩展\n",
    "合成特征\n",
    "特征组合\n",
    "特征交叉"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pickle\n",
    "import pandas as pd\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "# 载入数据\n",
    "with open('final.pkl', 'rb') as f:\n",
    "    final_data = pickle.load(f)\n",
    "# data = pd.read_csv(\"./data.csv\",encoding='gbk')\n",
    "# y=data.status\n",
    "# 划分训练集测试集\n",
    "X, y = final_data[final_data.columns.drop(\"status\")], final_data['status']\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=2018)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 110,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 性能评估\n",
    "from sklearn.metrics import accuracy_score, roc_auc_score\n",
    "\n",
    "def model_metrics(clf, X_train, X_test, y_train, y_test):\n",
    "    # 预测\n",
    "    y_train_pred = clf.predict(X_train)\n",
    "    y_test_pred = clf.predict(X_test)\n",
    "    \n",
    "    y_train_proba = clf.predict_proba(X_train)[:,1]\n",
    "    y_test_proba = clf.predict_proba(X_test)[:,1]\n",
    "    \n",
    "    # 准确率\n",
    "    print \"'[准确率]': end = ''\"\n",
    "    print \"训练集：', '%.4f', end = ' '\" % accuracy_score(y_train, y_train_pred)\n",
    "    print \"'测试集：', '%.4f')\" % accuracy_score(y_test, y_test_pred)\n",
    "    \n",
    "    # auc取值：用roc_auc_score或auc\n",
    "    print \"('[auc值]', end = ' ')\"\n",
    "    print \"('训练集：', '%.4f', end = ' ')\"%roc_auc_score(y_train, y_train_proba)\n",
    "    print \"'测试集：', '%.4f')\"%roc_auc_score(y_test, y_test_proba)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## IV值进行特征选择[参考链接](https://blog.csdn.net/iModel/article/details/79420437)\n",
    "\n",
    "在机器学习的二分类问题中，IV值（Information Value）主要用来对输入变量进行编码和预测能力评估。特征变量IV值的大小即表示该变量预测能力的强弱。IV 值的取值范围是 \\[[0, 正无穷) ，如果当前分组中只包含响应客户或者未响应客户时，IV = 正无穷。量化指标含义如下：< 0.02useless for prediction、0.02 to 0.1Weak predictor、0.1 to 0.3Medium predictor、0.3 to 0.5Strong predictor 、>0.5 Suspicious or too good to be true。\n",
    "\n",
    "- WOE的全称是“weight of evidence”，即证据权重。直观上讲，WOE是对原始变量的一种编码形式，要对一个变量进行WOE编码，首先需要把这个变量进行分组处理，即分箱或者离散化，常用离散化的方法有等宽分组，等高分组，或者利用决策树来分组。\n",
    "\n",
    "- IV衡量的是某一个变量的信息量，从公式来看的话，相当于是自变量WOE值的一个加权求和，其值的大小决定了自变量对于目标变量的影响程度，对于分组 i ，其对应的IV值参考下图，其中n是分组个数，注意，在变量的任何分组中，不应该出现响应数为0或非响应数位0的情况，当变量的一个分组的响应数位0时，对应的woe就为负无穷，此时IV值为正无穷。\n",
    "\n",
    "- WOE和IV值的区别WOE 和 IV 都能表达某个分组对目标变量的预测能力。但实际中，我们通常选择 IV 而不是 WOE 的和来衡量变量预测的能力，这是为什么呢？首先，因为我们在衡量一个变量的预测能力时，我们所使用的指标值不应该是负数。从这意义上来说，IV 比 WOE 多乘以前面那个因子，就保证了它不会是负数；然后，乘以(Pyi−Pni)这个因子，体现出了变量当前分组中个体的数量占整体的比例，从而很好考虑了这个分组中样本占整体的比例，比例越低，这个分组对变量整体预测能力的贡献越低。相反，如果直接用 WOE 的绝对值加和，会因为该分组出现次数偏少的影响而得到一个很高的指标。\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 111,
   "metadata": {},
   "outputs": [],
   "source": [
    "import math\n",
    "import numpy as np\n",
    "from scipy import stats\n",
    "from sklearn.utils.multiclass import type_of_target\n",
    "\n",
    "def woe(X, y, event=1):  \n",
    "    res_woe = []\n",
    "    iv_dict = {}\n",
    "    for feature in X.columns:\n",
    "        x = X[feature].values\n",
    "        # 1) 连续特征离散化\n",
    "        if type_of_target(x) == 'continuous':\n",
    "            x = discrete(x)\n",
    "        # 2) 计算该特征的woe和iv\n",
    "        # woe_dict, iv = woe_single_x(x, y, feature, event)\n",
    "        woe_dict, iv = woe_single_x(x, y, feature, event)\n",
    "        iv_dict[feature] = iv\n",
    "        res_woe.append(woe_dict) \n",
    "        \n",
    "    return iv_dict\n",
    "        \n",
    "def discrete(x):\n",
    "    # 使用5等分离散化特征\n",
    "    res = np.zeros(x.shape)\n",
    "    for i in range(5):\n",
    "        point1 = stats.scoreatpercentile(x, i * 20)\n",
    "        point2 = stats.scoreatpercentile(x, (i + 1) * 20)\n",
    "        x1 = x[np.where((x >= point1) & (x <= point2))]\n",
    "        mask = np.in1d(x, x1)\n",
    "        res[mask] = i + 1    # 将[i, i+1]块内的值标记成i+1\n",
    "    return res\n",
    "\n",
    "def woe_single_x(x, y, feature,event = 1):\n",
    "    # event代表预测正例的标签\n",
    "    event_total = sum(y == event)\n",
    "    non_event_total = y.shape[-1] - event_total\n",
    "    \n",
    "    iv = 0\n",
    "    woe_dict = {}\n",
    "    for x1 in set(x):    # 遍历各个块\n",
    "        y1 = y.reindex(np.where(x == x1)[0])\n",
    "        event_count = sum(y1 == event)\n",
    "        non_event_count = y1.shape[-1] - event_count\n",
    "        rate_event = event_count / event_total    \n",
    "        rate_non_event = non_event_count / non_event_total\n",
    "        \n",
    "        if rate_event == 0:\n",
    "            rate_event = 0.0001\n",
    "            # woei = -20\n",
    "        if rate_non_event == 0:\n",
    "            rate_non_event = 0.0001\n",
    "            # woei = 20\n",
    "        woei = math.log(rate_event / rate_non_event)\n",
    "        woe_dict[x1] = woei\n",
    "        iv += (rate_event - rate_non_event) * woei\n",
    "    return woe_dict, iv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "metadata": {},
   "outputs": [],
   "source": [
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\")\n",
    "\n",
    "iv_dict = woe(X_train, y_train)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 113,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[(u'reg_preference_for_trad_\\u4e8c\\u7ebf\\u57ce\\u5e02', 9.209419337938984),\n",
       " (u'reg_preference_for_trad_\\u5176\\u4ed6\\u57ce\\u5e02', 9.209419337938984),\n",
       " (u'is_high_user', 9.209419337938984),\n",
       " (u'railway_consume_count_last_12_month', 9.209419337938984),\n",
       " (u'reg_preference_for_trad_\\u5883\\u5916', 9.209419337938984),\n",
       " (u'jewelry_consume_count_last_6_month', 9.209419337938984),\n",
       " (u'loans_long_time', 0.0),\n",
       " (u'low_volume_percent', 0.0),\n",
       " (u'avg_price_last_12_month', 0.0),\n",
       " (u'avg_price_top_last_12_valid_month', 0.0),\n",
       " (u'latest_six_month_loan', 0.0),\n",
       " (u'consume_mini_time_last_1_month', 0.0),\n",
       " (u'latest_one_month_fail', 0.0),\n",
       " (u'consfin_product_count', 0.0),\n",
       " (u'trans_fail_top_count_enum_last_1_month', 0.0),\n",
       " (u'trans_day_last_12_month', 0.0),\n",
       " (u'middle_volume_percent', 0.0),\n",
       " (u'max_cumulative_consume_later_1_month', 0.0),\n",
       " (u'regional_mobility', 0.0),\n",
       " (u'consfin_credit_limit', 0.0),\n",
       " (u'trans_days_interval_filter', 0.0),\n",
       " (u'query_org_count', 0.0),\n",
       " (u'loans_score', 0.0),\n",
       " (u'loans_overdue_count', 0.0),\n",
       " (u'trans_amount_increase_rate_lately', 0.0),\n",
       " (u'query_sum_count', 0.0),\n",
       " (u'repayment_capability', 0.0),\n",
       " (u'first_transaction_day', 0.0),\n",
       " (u'consfin_org_count_current', 0.0),\n",
       " (u'transd_mcc', 0.0),\n",
       " (u'loans_product_count', 0.0),\n",
       " (u'take_amount_in_later_12_month_highest', 0.0),\n",
       " (u'reg_preference_for_trad_\\u4e09\\u7ebf\\u57ce\\u5e02', 0.0),\n",
       " (u'latest_one_month_apply', 0.0),\n",
       " (u'number_of_trans_from_2011', 0.0),\n",
       " (u'abs', 0.0),\n",
       " (u'loans_avg_limit', 0.0),\n",
       " (u'loans_org_count_current', 0.0),\n",
       " (u'loans_credit_limit', 0.0),\n",
       " (u'query_finance_count', 0.0),\n",
       " (u'apply_credibility', 0.0),\n",
       " (u'loans_count', 0.0),\n",
       " (u'rank_trad_1_month', 0.0),\n",
       " (u'consume_top_time_last_1_month', 0.0),\n",
       " (u'trans_fail_top_count_enum_last_6_month', 0.0),\n",
       " (u'historical_trans_amount', 0.0),\n",
       " (u'latest_three_month_loan', 0.0),\n",
       " (u'query_cash_count', 0.0),\n",
       " (u'loans_cash_count', 0.0),\n",
       " (u'consfin_max_limit', 0.0),\n",
       " (u'top_trans_count_last_1_month', 0.0),\n",
       " (u'pawns_auctions_trusts_consume_last_1_month', 0.0),\n",
       " (u'consfin_credibility', 0.0),\n",
       " (u'trans_fail_top_count_enum_last_12_month', 0.0),\n",
       " (u'trans_amount_3_month', 0.0),\n",
       " (u'latest_query_day', 0.0),\n",
       " (u'loans_latest_day', 0.0),\n",
       " (u'history_suc_fee', 0.0),\n",
       " (u'loans_credibility_behavior', 0.0),\n",
       " (u'consfin_avg_limit', 0.0),\n",
       " (u'loans_max_limit', 0.0),\n",
       " (u'latest_one_month_loan', 0.0),\n",
       " (u'history_fail_fee', 0.0),\n",
       " (u'historical_trans_day', 0.0),\n",
       " (u'latest_six_month_apply', 0.0),\n",
       " (u'loans_settle_count', 0.0),\n",
       " (u'latest_three_month_apply', 0.0),\n",
       " (u'avg_consume_less_12_valid_month', 0.0),\n",
       " (u'trans_activity_day', 0.0),\n",
       " (u'pawns_auctions_trusts_consume_last_6_month', 0.0),\n",
       " (u'student_feature', 0.0),\n",
       " (u'trans_days_interval', 0.0),\n",
       " (u'max_consume_count_later_6_month', 0.0),\n",
       " (u'loans_credibility_limit', 0.0),\n",
       " (u'loans_org_count_behavior', 0.0),\n",
       " (u'apply_score', 0.0),\n",
       " (u'consume_top_time_last_6_month', 0.0),\n",
       " (u'trans_top_time_last_1_month', 0.0),\n",
       " (u'trans_activity_month', 0.0),\n",
       " (u'reg_preference_for_trad_\\u4e00\\u7ebf\\u57ce\\u5e02', 0.0),\n",
       " (u'latest_one_month_suc', 0.0),\n",
       " (u'consfin_org_count_behavior', 0.0),\n",
       " (u'trans_top_time_last_6_month', 0.0),\n",
       " (u'cross_consume_count_last_1_month', 0.0)]"
      ]
     },
     "execution_count": 113,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "iv = sorted(iv_dict.items(), key = lambda x:x[1],reverse = True)\n",
    "iv"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 随机森林挑选特征"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 104,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "袋外分数： 0.7424105801021942\n",
      "'[准确率]': end = ''\n",
      "训练集：', '0.9850', end = ' '\n",
      "'测试集：', '0.7744')\n",
      "('[auc值]', end = ' ')\n",
      "('训练集：', '0.9992', end = ' ')\n",
      "'测试集：', '0.7193')\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',\n",
       "            max_depth=None, max_features='auto', max_leaf_nodes=None,\n",
       "            min_impurity_decrease=0.0, min_impurity_split=None,\n",
       "            min_samples_leaf=1, min_samples_split=2,\n",
       "            min_weight_fraction_leaf=0.0, n_estimators=10, n_jobs=1,\n",
       "            oob_score=True, random_state=2018, verbose=0, warm_start=False)"
      ]
     },
     "execution_count": 104,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\")\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "\n",
    "# 观察默认参数的性能\n",
    "rf0 = RandomForestClassifier(oob_score=True, random_state=2018)\n",
    "rf0.fit(X_train, y_train)\n",
    "print u'袋外分数：', rf0.oob_score_\n",
    "model_metrics(rf0, X_train, X_test, y_train, y_test)\n",
    "rf0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 105,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "({'n_estimators': 100}, 0.7939985894823949)"
      ]
     },
     "execution_count": 105,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 网格法调参\n",
    "param_test = {'n_estimators':range(20,200,20)}\n",
    "gsearch = GridSearchCV(estimator = RandomForestClassifier(n_estimators=120, max_depth=9, min_samples_split=50, \n",
    "                                                          min_samples_leaf=20, max_features = 9,random_state=2018), \n",
    "                       param_grid = param_test, scoring='roc_auc', cv=5)\n",
    "\n",
    "gsearch.fit(X_train, y_train)\n",
    "# gsearch.grid_scores_, \n",
    "gsearch.best_params_, gsearch.best_score_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "袋外分数： 0.7902013826269912\n",
      "'[准确率]': end = ''\n",
      "训练集：', '0.8182', end = ' '\n",
      "'测试集：', '0.7821')\n",
      "('[auc值]', end = ' ')\n",
      "('训练集：', '0.8999', end = ' ')\n",
      "'测试集：', '0.7728')\n"
     ]
    }
   ],
   "source": [
    "rf = RandomForestClassifier(n_estimators=100, max_depth=9, min_samples_split=50,\n",
    "                            min_samples_leaf=20, max_features = 9,oob_score=True, random_state=2018)\n",
    "rf.fit(X_train, y_train)\n",
    "print u'袋外分数：', rf.oob_score_\n",
    "model_metrics(rf, X_train, X_test, y_train, y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{u'abs': 0.006719667941365493,\n",
       " u'apply_credibility': 0.005117155285768405,\n",
       " u'apply_score': 0.06308268368322414,\n",
       " u'avg_consume_less_12_valid_month': 0.0015488547285952615,\n",
       " u'avg_price_last_12_month': 0.010072827889432335,\n",
       " u'avg_price_top_last_12_valid_month': 0.004634251845524828,\n",
       " u'consfin_avg_limit': 0.012167948971900702,\n",
       " u'consfin_credibility': 0.005578808638673587,\n",
       " u'consfin_credit_limit': 0.00894937059798237,\n",
       " u'consfin_max_limit': 0.0076301228627914455,\n",
       " u'consfin_org_count_behavior': 0.002491680778681817,\n",
       " u'consfin_org_count_current': 0.0041933816829697965,\n",
       " u'consfin_product_count': 0.004515194457619264,\n",
       " u'consume_mini_time_last_1_month': 0.006984134834602973,\n",
       " u'consume_top_time_last_1_month': 0.009705384040834536,\n",
       " u'consume_top_time_last_6_month': 0.006534534454032899,\n",
       " u'cross_consume_count_last_1_month': 0.00031588336700211044,\n",
       " u'first_transaction_day': 0.009583052272597613,\n",
       " u'historical_trans_amount': 0.010902055704639096,\n",
       " u'historical_trans_day': 0.008874509864511257,\n",
       " u'history_fail_fee': 0.1131460948020786,\n",
       " u'history_suc_fee': 0.01083947927223657,\n",
       " u'is_high_user': 0.0,\n",
       " u'jewelry_consume_count_last_6_month': 0.0,\n",
       " u'latest_one_month_apply': 0.0038745575655543575,\n",
       " u'latest_one_month_fail': 0.0500495827779603,\n",
       " u'latest_one_month_loan': 0.001728058090629679,\n",
       " u'latest_one_month_suc': 0.018943873520427777,\n",
       " u'latest_query_day': 0.017620250591160826,\n",
       " u'latest_six_month_apply': 0.004982926989015431,\n",
       " u'latest_six_month_loan': 0.005720427977752887,\n",
       " u'latest_three_month_apply': 0.006647329177593698,\n",
       " u'latest_three_month_loan': 0.006346436462771249,\n",
       " u'loans_avg_limit': 0.006679190805683137,\n",
       " u'loans_cash_count': 0.00408591381109072,\n",
       " u'loans_count': 0.008715199792197611,\n",
       " u'loans_credibility_behavior': 0.003017126828337721,\n",
       " u'loans_credibility_limit': 0.002400814867594533,\n",
       " u'loans_credit_limit': 0.003572253398359238,\n",
       " u'loans_latest_day': 0.009074808206172234,\n",
       " u'loans_long_time': 0.00788690141034075,\n",
       " u'loans_max_limit': 0.007478420420412541,\n",
       " u'loans_org_count_behavior': 0.003936060733122983,\n",
       " u'loans_org_count_current': 0.005218533778189632,\n",
       " u'loans_overdue_count': 0.04349707521896435,\n",
       " u'loans_product_count': 0.005093174794129803,\n",
       " u'loans_score': 0.07299907548161173,\n",
       " u'loans_settle_count': 0.009512221953146494,\n",
       " u'low_volume_percent': 0.0032812047153497562,\n",
       " u'max_consume_count_later_6_month': 0.0029857651904924816,\n",
       " u'max_cumulative_consume_later_1_month': 0.01970361477796049,\n",
       " u'middle_volume_percent': 0.005730455208392631,\n",
       " u'number_of_trans_from_2011': 0.005502641342458854,\n",
       " u'pawns_auctions_trusts_consume_last_1_month': 0.007414206175569922,\n",
       " u'pawns_auctions_trusts_consume_last_6_month': 0.007805559998288518,\n",
       " u'query_cash_count': 0.0025315467763918267,\n",
       " u'query_finance_count': 0.0031837315582433036,\n",
       " u'query_org_count': 0.003077664284070216,\n",
       " u'query_sum_count': 0.005037984773590202,\n",
       " u'railway_consume_count_last_12_month': 0.0,\n",
       " u'rank_trad_1_month': 0.016248946710905557,\n",
       " u'reg_preference_for_trad_\\u4e00\\u7ebf\\u57ce\\u5e02': 0.0013367206253297958,\n",
       " u'reg_preference_for_trad_\\u4e09\\u7ebf\\u57ce\\u5e02': 0.0005951234631504861,\n",
       " u'reg_preference_for_trad_\\u4e8c\\u7ebf\\u57ce\\u5e02': 0.0,\n",
       " u'reg_preference_for_trad_\\u5176\\u4ed6\\u57ce\\u5e02': 0.0,\n",
       " u'reg_preference_for_trad_\\u5883\\u5916': 0.00012305109617948365,\n",
       " u'regional_mobility': 0.0007591157763269282,\n",
       " u'repayment_capability': 0.00913813979851948,\n",
       " u'student_feature': 0.00046119592370538016,\n",
       " u'take_amount_in_later_12_month_highest': 0.005381039595138312,\n",
       " u'top_trans_count_last_1_month': 0.008826588444985866,\n",
       " u'trans_activity_day': 0.0100941566527476,\n",
       " u'trans_activity_month': 0.005503374857775672,\n",
       " u'trans_amount_3_month': 0.010883166394533912,\n",
       " u'trans_amount_increase_rate_lately': 0.007155896777015394,\n",
       " u'trans_day_last_12_month': 0.02237300997807261,\n",
       " u'trans_days_interval': 0.0052306851947874,\n",
       " u'trans_days_interval_filter': 0.008975015939729818,\n",
       " u'trans_fail_top_count_enum_last_12_month': 0.021850806333816367,\n",
       " u'trans_fail_top_count_enum_last_1_month': 0.12296704348417509,\n",
       " u'trans_fail_top_count_enum_last_6_month': 0.030280854961790782,\n",
       " u'trans_top_time_last_1_month': 0.014624721554774239,\n",
       " u'trans_top_time_last_6_month': 0.005470664136686101,\n",
       " u'transd_mcc': 0.0027990148697607338}"
      ]
     },
     "execution_count": 107,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rf_dict = {}\n",
    "feature_s = rf.feature_importances_\n",
    "feature_n = X.columns\n",
    "for i in range(0,len(feature_s)):\n",
    "    rf_dict[feature_n[i]] = feature_s[i]\n",
    "rf_dict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[(u'trans_fail_top_count_enum_last_1_month', 0.12296704348417509),\n",
       " (u'history_fail_fee', 0.1131460948020786),\n",
       " (u'loans_score', 0.07299907548161173),\n",
       " (u'apply_score', 0.06308268368322414),\n",
       " (u'latest_one_month_fail', 0.0500495827779603),\n",
       " (u'loans_overdue_count', 0.04349707521896435),\n",
       " (u'trans_fail_top_count_enum_last_6_month', 0.030280854961790782),\n",
       " (u'trans_day_last_12_month', 0.02237300997807261),\n",
       " (u'trans_fail_top_count_enum_last_12_month', 0.021850806333816367),\n",
       " (u'max_cumulative_consume_later_1_month', 0.01970361477796049),\n",
       " (u'latest_one_month_suc', 0.018943873520427777),\n",
       " (u'latest_query_day', 0.017620250591160826),\n",
       " (u'rank_trad_1_month', 0.016248946710905557),\n",
       " (u'trans_top_time_last_1_month', 0.014624721554774239),\n",
       " (u'consfin_avg_limit', 0.012167948971900702),\n",
       " (u'historical_trans_amount', 0.010902055704639096),\n",
       " (u'trans_amount_3_month', 0.010883166394533912),\n",
       " (u'history_suc_fee', 0.01083947927223657),\n",
       " (u'trans_activity_day', 0.0100941566527476),\n",
       " (u'avg_price_last_12_month', 0.010072827889432335),\n",
       " (u'consume_top_time_last_1_month', 0.009705384040834536),\n",
       " (u'first_transaction_day', 0.009583052272597613),\n",
       " (u'loans_settle_count', 0.009512221953146494),\n",
       " (u'repayment_capability', 0.00913813979851948),\n",
       " (u'loans_latest_day', 0.009074808206172234),\n",
       " (u'trans_days_interval_filter', 0.008975015939729818),\n",
       " (u'consfin_credit_limit', 0.00894937059798237),\n",
       " (u'historical_trans_day', 0.008874509864511257),\n",
       " (u'top_trans_count_last_1_month', 0.008826588444985866),\n",
       " (u'loans_count', 0.008715199792197611),\n",
       " (u'loans_long_time', 0.00788690141034075),\n",
       " (u'pawns_auctions_trusts_consume_last_6_month', 0.007805559998288518),\n",
       " (u'consfin_max_limit', 0.0076301228627914455),\n",
       " (u'loans_max_limit', 0.007478420420412541),\n",
       " (u'pawns_auctions_trusts_consume_last_1_month', 0.007414206175569922),\n",
       " (u'trans_amount_increase_rate_lately', 0.007155896777015394),\n",
       " (u'consume_mini_time_last_1_month', 0.006984134834602973),\n",
       " (u'abs', 0.006719667941365493),\n",
       " (u'loans_avg_limit', 0.006679190805683137),\n",
       " (u'latest_three_month_apply', 0.006647329177593698),\n",
       " (u'consume_top_time_last_6_month', 0.006534534454032899),\n",
       " (u'latest_three_month_loan', 0.006346436462771249),\n",
       " (u'middle_volume_percent', 0.005730455208392631),\n",
       " (u'latest_six_month_loan', 0.005720427977752887),\n",
       " (u'consfin_credibility', 0.005578808638673587),\n",
       " (u'trans_activity_month', 0.005503374857775672),\n",
       " (u'number_of_trans_from_2011', 0.005502641342458854),\n",
       " (u'trans_top_time_last_6_month', 0.005470664136686101),\n",
       " (u'take_amount_in_later_12_month_highest', 0.005381039595138312),\n",
       " (u'trans_days_interval', 0.0052306851947874),\n",
       " (u'loans_org_count_current', 0.005218533778189632),\n",
       " (u'apply_credibility', 0.005117155285768405),\n",
       " (u'loans_product_count', 0.005093174794129803),\n",
       " (u'query_sum_count', 0.005037984773590202),\n",
       " (u'latest_six_month_apply', 0.004982926989015431),\n",
       " (u'avg_price_top_last_12_valid_month', 0.004634251845524828),\n",
       " (u'consfin_product_count', 0.004515194457619264),\n",
       " (u'consfin_org_count_current', 0.0041933816829697965),\n",
       " (u'loans_cash_count', 0.00408591381109072),\n",
       " (u'loans_org_count_behavior', 0.003936060733122983),\n",
       " (u'latest_one_month_apply', 0.0038745575655543575),\n",
       " (u'loans_credit_limit', 0.003572253398359238),\n",
       " (u'low_volume_percent', 0.0032812047153497562),\n",
       " (u'query_finance_count', 0.0031837315582433036),\n",
       " (u'query_org_count', 0.003077664284070216),\n",
       " (u'loans_credibility_behavior', 0.003017126828337721),\n",
       " (u'max_consume_count_later_6_month', 0.0029857651904924816),\n",
       " (u'transd_mcc', 0.0027990148697607338),\n",
       " (u'query_cash_count', 0.0025315467763918267),\n",
       " (u'consfin_org_count_behavior', 0.002491680778681817),\n",
       " (u'loans_credibility_limit', 0.002400814867594533),\n",
       " (u'latest_one_month_loan', 0.001728058090629679),\n",
       " (u'avg_consume_less_12_valid_month', 0.0015488547285952615),\n",
       " (u'reg_preference_for_trad_\\u4e00\\u7ebf\\u57ce\\u5e02', 0.0013367206253297958),\n",
       " (u'regional_mobility', 0.0007591157763269282),\n",
       " (u'reg_preference_for_trad_\\u4e09\\u7ebf\\u57ce\\u5e02', 0.0005951234631504861),\n",
       " (u'student_feature', 0.00046119592370538016),\n",
       " (u'cross_consume_count_last_1_month', 0.00031588336700211044),\n",
       " (u'reg_preference_for_trad_\\u5883\\u5916', 0.00012305109617948365),\n",
       " (u'reg_preference_for_trad_\\u4e8c\\u7ebf\\u57ce\\u5e02', 0.0),\n",
       " (u'reg_preference_for_trad_\\u5176\\u4ed6\\u57ce\\u5e02', 0.0),\n",
       " (u'is_high_user', 0.0),\n",
       " (u'railway_consume_count_last_12_month', 0.0),\n",
       " (u'jewelry_consume_count_last_6_month', 0.0)]"
      ]
     },
     "execution_count": 108,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rf_sort = sorted(rf_dict.items(), key = lambda x:x[1],reverse = True)\n",
    "rf_sort"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.15"
  },
  "toc": {
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": true,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}