experiment_tjh.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.\n",
      "2023-05-04 22:36:26.339758: I tensorflow/core/platform/cpu_feature_guard.cc:193] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  AVX2 AVX512F AVX512_VNNI FMA\n",
      "To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.\n",
      "2023-05-04 22:36:26.484198: I tensorflow/core/util/port.cc:104] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.\n",
      "2023-05-04 22:36:27.282441: W tensorflow/compiler/xla/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libnvinfer.so.7'; dlerror: libnvinfer.so.7: cannot open shared object file: No such file or directory; LD_LIBRARY_PATH: /home/ubuntu/.local/lib/python3.9/site-packages/cv2/../../lib64::/usr/local/cuda/lib64\n",
      "2023-05-04 22:36:27.282606: W tensorflow/compiler/xla/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libnvinfer_plugin.so.7'; dlerror: libnvinfer_plugin.so.7: cannot open shared object file: No such file or directory; LD_LIBRARY_PATH: /home/ubuntu/.local/lib/python3.9/site-packages/cv2/../../lib64::/usr/local/cuda/lib64\n",
      "2023-05-04 22:36:27.282618: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Cannot dlopen some TensorRT libraries. If you would like to use Nvidia GPU with TensorRT, please make sure the missing libraries mentioned above are installed properly.\n"
     ]
    }
   ],
   "source": [
    "from aif360.algorithms.preprocessing import Reweighing\n",
    "# Fairness metrics\n",
    "from aif360.metrics import BinaryLabelDatasetMetric\n",
    "from aif360.metrics import ClassificationMetric\n",
    "from collections import defaultdict\n",
    "from methods.mitigate_disparity import MultiLevelReweighing\n",
    "import numpy as np\n",
    "\n",
    "from sklearn.pipeline import make_pipeline\n",
    "from metrics.eval_metrics import print_metrics_binary\n",
    "\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "\n",
    "from aif360.datasets import StandardDataset\n",
    "import pandas as pd\n",
    "np.random.seed(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 评价数据集\n",
    "def test(dataset, model, thresh_arr,unprivileged_groups,privileged_groups):\n",
    "    try:\n",
    "        # sklearn classifier\n",
    "        # 得到一个n行2列的概率矩阵，每行两个概率和为1\n",
    "        y_val_pred_prob = model.predict_proba(dataset.features)\n",
    "        # 得到favorable_label对应位置\n",
    "        pos_ind = np.where(model.classes_ == dataset.favorable_label)[0][0]\n",
    "    except AttributeError:\n",
    "        # aif360 inprocessing algorithm\n",
    "        y_val_pred_prob = model.predict(dataset).scores\n",
    "        pos_ind = 0\n",
    "    #创建一个dict，其中每个value是一个list\n",
    "    metric_arrs = defaultdict(list)\n",
    "    # 这里的thresh_arr为等间隔区间\n",
    "    for thresh in thresh_arr:\n",
    "        # 阈值筛选，对应概率大于阈值为1，小于为0\n",
    "        y_val_pred = (y_val_pred_prob[:, pos_ind] > thresh).astype(np.float64)\n",
    "\n",
    "        dataset_pred = dataset.copy()\n",
    "        dataset_pred.labels = y_val_pred\n",
    "        # 生成计算指标\n",
    "        metric = ClassificationMetric(\n",
    "                dataset, dataset_pred,\n",
    "                unprivileged_groups=unprivileged_groups,\n",
    "                privileged_groups=privileged_groups)\n",
    "        acc_metrics_binary = print_metrics_binary(dataset.labels.ravel(),y_val_pred)\n",
    "        # 添加指标属性\n",
    "        metric_arrs['acc'].append(acc_metrics_binary['acc'])\n",
    "        metric_arrs['auroc'].append(acc_metrics_binary['auroc'])\n",
    "        metric_arrs['auprc'].append(acc_metrics_binary['auprc'])\n",
    "        metric_arrs['bal_acc'].append((metric.true_positive_rate()\n",
    "                                     + metric.true_negative_rate()) / 2)\n",
    "        metric_arrs['avg_odds_diff'].append(metric.average_odds_difference())\n",
    "        metric_arrs['disp_imp'].append(metric.disparate_impact())\n",
    "        metric_arrs['stat_par_diff'].append(metric.statistical_parity_difference())\n",
    "        metric_arrs['eq_opp_diff'].append(metric.equal_opportunity_difference())\n",
    "        metric_arrs['theil_ind'].append(metric.theil_index())\n",
    "    \n",
    "    return metric_arrs\n",
    "def describe_metrics(metrics, thresh_arr):\n",
    "    best_ind = np.argmax(metrics['bal_acc'])\n",
    "    print(\"Threshold corresponding to Best balanced accuracy: {:6.4f}\".format(thresh_arr[best_ind]))\n",
    "    print(\"Best balanced accuracy: {:6.4f}\".format(metrics['bal_acc'][best_ind]))\n",
    "    print(\"acc value: {:6.4f}\".format(metrics['acc'][best_ind]))\n",
    "    print(\"auroc value: {:6.4f}\".format(metrics['auroc'][best_ind]))\n",
    "    print(\"auprc value: {:6.4f}\".format(metrics['auprc'][best_ind]))\n",
    "#     disp_imp_at_best_ind = np.abs(1 - np.array(metrics['disp_imp']))[best_ind]\n",
    "    disp_imp_at_best_ind = min(metrics['disp_imp'][best_ind], 1/metrics['disp_imp'][best_ind])\n",
    "    print(\"Corresponding min(DI, 1/DI) value: {:6.4f}\".format(disp_imp_at_best_ind))\n",
    "    print(\"Corresponding average odds difference value: {:6.4f}\".format(metrics['avg_odds_diff'][best_ind]))\n",
    "    print(\"Corresponding statistical parity difference value: {:6.4f}\".format(metrics['stat_par_diff'][best_ind]))\n",
    "    print(\"Corresponding equal opportunity difference value: {:6.4f}\".format(metrics['eq_opp_diff'][best_ind]))\n",
    "    print(\"Corresponding Theil index value: {:6.4f}\".format(metrics['theil_ind'][best_ind]))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "feature_name_level_dic={'age': 2, 'gender': 1}\n",
    "data_df = pd.read_csv(\n",
    "    './datasets/tjh.csv')\n",
    "grouped_data = data_df.groupby('patiend_id')\n",
    "data_first_visit = grouped_data.first().reset_index()\n",
    "# data_first_visit=data_df\n",
    "data_first_visit.drop('patiend_id', axis=1, inplace=True)\n",
    "protect_feature_list = list(feature_name_level_dic.keys())\n",
    "privileged_classes = []\n",
    "protected_attribute_names = []\n",
    "for protect_feature_index in range(len(protect_feature_list)):\n",
    "    protect_feature_name = protect_feature_list[protect_feature_index]\n",
    "    avg = data_first_visit[protect_feature_name].mean()\n",
    "    data_first_visit[protect_feature_name] = (\n",
    "        data_first_visit[protect_feature_name] > avg).astype(int)\n",
    "    privileged_classes.append([1.0])\n",
    "    protected_attribute_names.append(protect_feature_name)\n",
    "\n",
    "dataset = StandardDataset(df=data_first_visit, label_name='outcome',\n",
    "                            favorable_classes=[1.0],\n",
    "                            protected_attribute_names=protected_attribute_names,\n",
    "                            privileged_classes=privileged_classes)\n",
    "(dataset_orig_train,\n",
    " dataset_orig_val) = dataset.split([0.7], shuffle=True)\n",
    "multi_privileged_groups = [\n",
    "    {\"feature_name\": \"age\", \"privileged_value\": 1, \"level\": 2},\n",
    "    {\"feature_name\": \"gender\", \"privileged_value\": 1, \"level\": 1},\n",
    "]\n",
    "multi_unprivileged_groups = [\n",
    "    {\"feature_name\": \"age\", \"unprivileged_value\": 0, \"level\": 2},\n",
    "    {\"feature_name\": \"gender\", \"unprivileged_value\": 0, \"level\": 1},\n",
    "]\n",
    "privileged_groups1 = [{\"age\": 1}]\n",
    "unprivileged_groups1 = [{\"age\": 0}]\n",
    "privileged_groups2 = [{\"gender\": 1}]\n",
    "unprivileged_groups2 = [{\"gender\": 0}]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "before reweighing ,age disparate impact and spd are 0.27816769605090774 -0.4823952470293934\n",
      "before reweighing ,gender disparate impact are 0.545413870246085 -0.2573129036342915\n",
      "age:\n",
      "Threshold corresponding to Best balanced accuracy: 0.2900\n",
      "Best balanced accuracy: 0.8357\n",
      "acc value: 0.8349\n",
      "auroc value: 0.8357\n",
      "auprc value: 0.8696\n",
      "Corresponding min(DI, 1/DI) value: 0.3147\n",
      "Corresponding average odds difference value: -0.2585\n",
      "Corresponding statistical parity difference value: -0.5227\n",
      "Corresponding equal opportunity difference value: -0.2922\n",
      "Corresponding Theil index value: 0.0989\n",
      "gender:\n",
      "Threshold corresponding to Best balanced accuracy: 0.2900\n",
      "Best balanced accuracy: 0.8357\n",
      "acc value: 0.8349\n",
      "auroc value: 0.8357\n",
      "auprc value: 0.8696\n",
      "Corresponding min(DI, 1/DI) value: 0.6732\n",
      "Corresponding average odds difference value: -0.0699\n",
      "Corresponding statistical parity difference value: -0.1942\n",
      "Corresponding equal opportunity difference value: -0.0018\n",
      "Corresponding Theil index value: 0.0989\n"
     ]
    }
   ],
   "source": [
    "ori_age = BinaryLabelDatasetMetric(dataset=dataset,unprivileged_groups=unprivileged_groups1,privileged_groups=privileged_groups1)\n",
    "ori_gender = BinaryLabelDatasetMetric(dataset=dataset,unprivileged_groups=unprivileged_groups2,privileged_groups=privileged_groups2)\n",
    "print(\n",
    "    \"before reweighing ,age disparate impact and spd are \"\n",
    "    + str(ori_age.disparate_impact())+\" \"+str(ori_age.statistical_parity_difference())\n",
    "\n",
    ")\n",
    "print(\n",
    "    \"before reweighing ,gender disparate impact are \"\n",
    "    + str(ori_gender.disparate_impact())+\" \"+str(ori_gender.statistical_parity_difference())\n",
    ")\n",
    "# 逻辑回归的pipeline\n",
    "model = make_pipeline(LogisticRegression(solver='liblinear', random_state=1))\n",
    "fit_params = {\n",
    "    'logisticregression__sample_weight': dataset_orig_train.instance_weights}\n",
    "# 原始训练数据逻辑回归\n",
    "lr_orig = model.fit(dataset_orig_train.features,\n",
    "                    dataset_orig_train.labels.ravel(), **fit_params)\n",
    "thresh_arr = np.linspace(0.01, 0.5, 50)\n",
    "# 评价验证数据逻辑回归\n",
    "print(\"age:\")\n",
    "val_metrics = test(dataset=dataset_orig_val,\n",
    "                   model=lr_orig,\n",
    "                   thresh_arr=thresh_arr,\n",
    "                   unprivileged_groups=unprivileged_groups1, privileged_groups=privileged_groups1)\n",
    "\n",
    "describe_metrics(val_metrics, thresh_arr)\n",
    "print(\"gender:\")\n",
    "val_metrics = test(dataset=dataset_orig_val,\n",
    "                   model=lr_orig,\n",
    "                   thresh_arr=thresh_arr,\n",
    "                   unprivileged_groups=unprivileged_groups2, privileged_groups=privileged_groups2)\n",
    "\n",
    "describe_metrics(val_metrics, thresh_arr)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "after reweighing ,age disparate impact and spd are 1.0 0.0\n",
      "after reweighing ,gender disparate impact and spd are 0.9999999999999991 -3.885780586188048e-16\n",
      "sex:\n",
      "Threshold corresponding to Best balanced accuracy: 0.1500\n",
      "Best balanced accuracy: 0.8095\n",
      "acc value: 0.8073\n",
      "auroc value: 0.8095\n",
      "auprc value: 0.8500\n",
      "Corresponding min(DI, 1/DI) value: 0.4104\n",
      "Corresponding average odds difference value: -0.2191\n",
      "Corresponding statistical parity difference value: -0.4597\n",
      "Corresponding equal opportunity difference value: -0.3160\n",
      "Corresponding Theil index value: 0.0969\n",
      "race:\n",
      "Threshold corresponding to Best balanced accuracy: 0.1500\n",
      "Best balanced accuracy: 0.8095\n",
      "acc value: 0.8073\n",
      "auroc value: 0.8095\n",
      "auprc value: 0.8500\n",
      "Corresponding min(DI, 1/DI) value: 0.7622\n",
      "Corresponding average odds difference value: -0.0392\n",
      "Corresponding statistical parity difference value: -0.1482\n",
      "Corresponding equal opportunity difference value: -0.0281\n",
      "Corresponding Theil index value: 0.0969\n"
     ]
    }
   ],
   "source": [
    "mmrw = MultiLevelReweighing(multi_unprivileged_groups, multi_privileged_groups)\n",
    "trans_dataset = mmrw.fit(dataset).transform(dataset)\n",
    "trans_dataset_train = mmrw.fit(dataset_orig_train).transform(dataset_orig_train)\n",
    "# 给定测量数据集的指标\n",
    "trans_age = BinaryLabelDatasetMetric(dataset=trans_dataset_train,unprivileged_groups=unprivileged_groups1,privileged_groups=privileged_groups1)\n",
    "trans_gender = BinaryLabelDatasetMetric(dataset=trans_dataset_train,unprivileged_groups=unprivileged_groups2,privileged_groups=privileged_groups2)\n",
    "\n",
    "\n",
    "print(\n",
    "    \"after reweighing ,age disparate impact and spd are \"\n",
    "\n",
    "    + str(trans_age.disparate_impact())+\" \"\n",
    "    + str(trans_age.statistical_parity_difference())\n",
    ")\n",
    "\n",
    "\n",
    "print(\n",
    "    \"after reweighing ,gender disparate impact and spd are \"\n",
    "\n",
    "    + str(trans_gender.disparate_impact())+\" \"\n",
    "    + str(trans_gender.statistical_parity_difference())\n",
    ")\n",
    "\n",
    "# 逻辑回归的pipeline\n",
    "model = make_pipeline(LogisticRegression(solver='liblinear', random_state=1))\n",
    "fit_params = {\n",
    "    'logisticregression__sample_weight': trans_dataset_train.instance_weights}\n",
    "# 原始训练数据逻辑回归\n",
    "lr_mmrw = model.fit(trans_dataset_train.features,\n",
    "                    trans_dataset_train.labels.ravel(), **fit_params)\n",
    "thresh_arr = np.linspace(0.01, 0.5, 50)\n",
    "# 评价验证数据逻辑回归\n",
    "print(\"sex:\")\n",
    "val_metrics = test(dataset=dataset_orig_val,\n",
    "                   model=lr_mmrw,\n",
    "                   thresh_arr=thresh_arr,\n",
    "                   unprivileged_groups=unprivileged_groups1, privileged_groups=privileged_groups1)\n",
    "lr_orig_best_ind = np.argmax(val_metrics['bal_acc'])\n",
    "describe_metrics(val_metrics, thresh_arr)\n",
    "print(\"race:\")\n",
    "val_metrics = test(dataset=dataset_orig_val,\n",
    "                   model=lr_mmrw,\n",
    "                   thresh_arr=thresh_arr,\n",
    "                   unprivileged_groups=unprivileged_groups2, privileged_groups=privileged_groups2)\n",
    "lr_orig_best_ind = np.argmax(val_metrics['bal_acc'])\n",
    "describe_metrics(val_metrics, thresh_arr)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "after reweighing ,age disparate impact and spd are 1.010969596827694 0.005020922431002661\n",
      "after reweighing ,gender disparate impact and spd are 0.9999999999999997 -1.6653345369377348e-16\n",
      "age:\n",
      "Threshold corresponding to Best balanced accuracy: 0.1600\n",
      "Best balanced accuracy: 0.8095\n",
      "acc value: 0.8073\n",
      "auroc value: 0.8095\n",
      "auprc value: 0.8500\n",
      "Corresponding min(DI, 1/DI) value: 0.4104\n",
      "Corresponding average odds difference value: -0.2191\n",
      "Corresponding statistical parity difference value: -0.4597\n",
      "Corresponding equal opportunity difference value: -0.3160\n",
      "Corresponding Theil index value: 0.0969\n",
      "gender:\n",
      "Threshold corresponding to Best balanced accuracy: 0.1600\n",
      "Best balanced accuracy: 0.8095\n",
      "acc value: 0.8073\n",
      "auroc value: 0.8095\n",
      "auprc value: 0.8500\n",
      "Corresponding min(DI, 1/DI) value: 0.7622\n",
      "Corresponding average odds difference value: -0.0392\n",
      "Corresponding statistical parity difference value: -0.1482\n",
      "Corresponding equal opportunity difference value: -0.0281\n",
      "Corresponding Theil index value: 0.0969\n"
     ]
    }
   ],
   "source": [
    "rw_age = Reweighing(unprivileged_groups=unprivileged_groups1,\n",
    "                    privileged_groups=privileged_groups1)\n",
    "rw_gender = Reweighing(unprivileged_groups=unprivileged_groups2,\n",
    "                     privileged_groups=privileged_groups2)\n",
    "\n",
    "trans_age_dataset = rw_age.fit(dataset).transform(dataset)\n",
    "trans_age_gender_dataset = rw_gender.fit(\n",
    "    trans_age_dataset).transform(trans_age_dataset)\n",
    "\n",
    "trans_age_dataset_train = rw_age.fit(dataset_orig_train).transform(dataset_orig_train)\n",
    "trans_age_gender_dataset_train = rw_gender.fit(\n",
    "    trans_age_dataset_train).transform(trans_age_dataset_train)\n",
    "\n",
    "trans_age_metric = BinaryLabelDatasetMetric(\n",
    "    dataset=trans_age_gender_dataset, unprivileged_groups=unprivileged_groups1, privileged_groups=privileged_groups1)\n",
    "trans_gender_metric = BinaryLabelDatasetMetric(\n",
    "    dataset=trans_age_gender_dataset, unprivileged_groups=unprivileged_groups2, privileged_groups=privileged_groups2)\n",
    "\n",
    "print(\n",
    "    \"after reweighing ,age disparate impact and spd are \"\n",
    "\n",
    "    + str(trans_age_metric.disparate_impact())+\" \"\n",
    "    + str(trans_age_metric.statistical_parity_difference())\n",
    ")\n",
    "\n",
    "\n",
    "print(\n",
    "    \"after reweighing ,gender disparate impact and spd are \"\n",
    "\n",
    "    + str(trans_gender_metric.disparate_impact())+\" \"\n",
    "    + str(trans_gender_metric.statistical_parity_difference())\n",
    ")\n",
    "# 逻辑回归的pipeline\n",
    "model = make_pipeline(LogisticRegression(solver='liblinear', random_state=1))\n",
    "fit_params = {\n",
    "    'logisticregression__sample_weight': trans_age_gender_dataset_train.instance_weights}\n",
    "# 原始训练数据逻辑回归\n",
    "lr_rw = model.fit(trans_age_gender_dataset_train.features,\n",
    "                    trans_age_gender_dataset_train.labels.ravel(), **fit_params)\n",
    "thresh_arr = np.linspace(0.01, 0.5, 50)\n",
    "# 评价验证数据逻辑回归\n",
    "print(\"age:\")\n",
    "val_metrics = test(dataset=dataset_orig_val,\n",
    "                   model=lr_rw,\n",
    "                   thresh_arr=thresh_arr,\n",
    "                   unprivileged_groups=unprivileged_groups1, privileged_groups=privileged_groups1)\n",
    "\n",
    "describe_metrics(val_metrics, thresh_arr)\n",
    "print(\"gender:\")\n",
    "val_metrics = test(dataset=dataset_orig_val,\n",
    "                   model=lr_rw,\n",
    "                   thresh_arr=thresh_arr,\n",
    "                   unprivileged_groups=unprivileged_groups2, privileged_groups=privileged_groups2)\n",
    "\n",
    "describe_metrics(val_metrics, thresh_arr)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "after reweighing ,age disparate impact and spd are 0.9999999999999999 -5.551115123125783e-17\n",
      "after reweighing ,gender disparate impact and spd are 1.0214939048830591 0.009800601436590783\n",
      "age:\n",
      "Threshold corresponding to Best balanced accuracy: 0.1600\n",
      "Best balanced accuracy: 0.8095\n",
      "acc value: 0.8073\n",
      "auroc value: 0.8095\n",
      "auprc value: 0.8500\n",
      "Corresponding min(DI, 1/DI) value: 0.4104\n",
      "Corresponding average odds difference value: -0.2191\n",
      "Corresponding statistical parity difference value: -0.4597\n",
      "Corresponding equal opportunity difference value: -0.3160\n",
      "Corresponding Theil index value: 0.0969\n",
      "gender:\n",
      "Threshold corresponding to Best balanced accuracy: 0.1600\n",
      "Best balanced accuracy: 0.8095\n",
      "acc value: 0.8073\n",
      "auroc value: 0.8095\n",
      "auprc value: 0.8500\n",
      "Corresponding min(DI, 1/DI) value: 0.7622\n",
      "Corresponding average odds difference value: -0.0392\n",
      "Corresponding statistical parity difference value: -0.1482\n",
      "Corresponding equal opportunity difference value: -0.0281\n",
      "Corresponding Theil index value: 0.0969\n"
     ]
    }
   ],
   "source": [
    "rw_age = Reweighing(unprivileged_groups=unprivileged_groups1,\n",
    "                    privileged_groups=privileged_groups1)\n",
    "rw_gender = Reweighing(unprivileged_groups=unprivileged_groups2,\n",
    "                     privileged_groups=privileged_groups2)\n",
    "trans_gender_dataset = rw_gender.fit(\n",
    "    dataset).transform(dataset)\n",
    "trans_gender_age_dataset = rw_age.fit(trans_gender_dataset).transform(trans_gender_dataset)\n",
    "\n",
    "\n",
    "trans_gender_dataset_train = rw_gender.fit(dataset_orig_train).transform(dataset_orig_train)\n",
    "trans_gender_age_dataset_train = rw_age.fit(\n",
    "    trans_gender_dataset_train).transform(trans_gender_dataset_train)\n",
    "\n",
    "trans_age_metric2 = BinaryLabelDatasetMetric(\n",
    "    dataset=trans_gender_age_dataset, unprivileged_groups=unprivileged_groups1, privileged_groups=privileged_groups1)\n",
    "trans_gender_metric2 = BinaryLabelDatasetMetric(\n",
    "    dataset=trans_gender_age_dataset, unprivileged_groups=unprivileged_groups2, privileged_groups=privileged_groups2)\n",
    "\n",
    "print(\n",
    "    \"after reweighing ,age disparate impact and spd are \"\n",
    "\n",
    "    + str(trans_age_metric2.disparate_impact())+\" \"\n",
    "    + str(trans_age_metric2.statistical_parity_difference())\n",
    ")\n",
    "\n",
    "\n",
    "print(\n",
    "    \"after reweighing ,gender disparate impact and spd are \"\n",
    "\n",
    "    + str(trans_gender_metric2.disparate_impact())+\" \"\n",
    "    + str(trans_gender_metric2.statistical_parity_difference())\n",
    ")\n",
    "# 逻辑回归的pipeline\n",
    "model = make_pipeline(LogisticRegression(solver='liblinear', random_state=1))\n",
    "fit_params = {\n",
    "    'logisticregression__sample_weight': trans_gender_age_dataset_train.instance_weights}\n",
    "# 原始训练数据逻辑回归\n",
    "lr_rw = model.fit(trans_gender_age_dataset_train.features,\n",
    "                    trans_gender_age_dataset_train.labels.ravel(), **fit_params)\n",
    "thresh_arr = np.linspace(0.01, 0.5, 50)\n",
    "# 评价验证数据逻辑回归\n",
    "print(\"age:\")\n",
    "val_metrics = test(dataset=dataset_orig_val,\n",
    "                   model=lr_rw,\n",
    "                   thresh_arr=thresh_arr,\n",
    "                   unprivileged_groups=unprivileged_groups1, privileged_groups=privileged_groups1)\n",
    "\n",
    "describe_metrics(val_metrics, thresh_arr)\n",
    "print(\"gender:\")\n",
    "val_metrics = test(dataset=dataset_orig_val,\n",
    "                   model=lr_rw,\n",
    "                   thresh_arr=thresh_arr,\n",
    "                   unprivileged_groups=unprivileged_groups2, privileged_groups=privileged_groups2)\n",
    "\n",
    "describe_metrics(val_metrics, thresh_arr)\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "med",
   "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": "3.9.15"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}