initial commit

.gitignore

+.ipynb_checkpoints
+__pycache__
\ No newline at end of file

1. Risk Explanation.ipynb

+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "> 在这个notebook中，我们测试UQER与TF的风险模型方差解释度"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "%matplotlib inline\n",
+    "import numpy as np\n",
+    "import pandas as pd\n",
+    "from tf.CovCal import RiskMatrics\n",
+    "from tf.CovCal import BarraCov\n",
+    "from PyFin.api import advanceDateByCalendar"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## 0. Data Preparation\n",
+    "------------------"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "risk_cov = pd.read_parquet('data/risk_cov.parqut').sort_values(['trade_date', 'FactorID']).set_index('trade_date')\n",
+    "risk_exp = pd.read_parquet('data/risk_exposure.parqut').sort_values(['trade_date', 'code']).set_index('trade_date')\n",
+    "specific_risk = pd.read_parquet('data/special_risk.parqut').sort_values(['trade_date', 'code']).set_index('trade_date')\n",
+    "risk_return = pd.read_parquet('data/risk_return.parqut').sort_values('trade_date').set_index('trade_date')"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "risk_factors = risk_cov.columns[2:-1].tolist()\n",
+    "dates = risk_cov.index.unique().tolist()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## 1. UQER v.s. TF Risk Model\n",
+    "--------------------"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "res = pd.DataFrame(columns=['UQER', 'TF', 'avg. vol', '# of Stk.'])\n",
+    "\n",
+    "window = '250b'\n",
+    "\n",
+    "for sample_date in dates[250:]:\n",
+    "    \n",
+    "    ## Uqer Risk Model\n",
+    "    sample_risk_cov = risk_cov.loc[sample_date:sample_date, risk_factors]\n",
+    "    sample_risk_exp = risk_exp.loc[sample_date:sample_date, ['code'] + risk_factors]\n",
+    "    sample_specific_risk = specific_risk.loc[sample_date:sample_date, ['code', 'SRISK']]\n",
+    "\n",
+    "    sample_codes = set(sample_risk_exp.code).intersection(sample_specific_risk.code)\n",
+    "    sample_risk_exp = sample_risk_exp.loc[sample_risk_exp.code.isin(sample_codes)]\n",
+    "    sample_specific_risk = sample_specific_risk.loc[sample_specific_risk.code.isin(sample_codes)]\n",
+    "\n",
+    "    sample_risk_exp = sample_risk_exp[risk_factors].values\n",
+    "    sample_risk_cov = sample_risk_cov.values\n",
+    "    sample_specific_risk = sample_specific_risk['SRISK'].values\n",
+    "    \n",
+    "    explained_cov = sample_risk_exp @ sample_risk_cov @ sample_risk_exp.T / 10000.\n",
+    "    sec_cov = explained_cov + np.diag(sample_specific_risk ** 2) / 10000\n",
+    "    \n",
+    "    explained_var = np.diag(explained_cov)\n",
+    "    sec_var = np.diag(sec_cov)\n",
+    "    explained_per = np.mean(explained_var / sec_var)\n",
+    "    avg_vol = np.mean(np.sqrt(sec_var))\n",
+    "    num_stks = len(sample_codes)\n",
+    "    \n",
+    "    ## TF Risk Model\n",
+    "    start_date = advanceDateByCalendar('china.sse', sample_date, period= '-' + window)\n",
+    "    sample_risk_return = risk_return[start_date:sample_date]\n",
+    "    ret_values = sample_risk_return[risk_factors].T\n",
+    "    tf_risk_cov = BarraCov(ret_values, 90, 20) * 10000 * 250\n",
+    "    tf_explained_cov = sample_risk_exp @ tf_risk_cov @ sample_risk_exp.T / 10000.\n",
+    "    tf_explained_var = np.diag(tf_explained_cov)\n",
+    "    tf_explained_per = np.mean(tf_explained_var / sec_var)\n",
+    "    \n",
+    "    ## Output\n",
+    "    res.loc[sample_date, 'UQER'] = explained_per\n",
+    "    res.loc[sample_date, 'UQER (std.)'] = np.std(explained_var / sec_var)\n",
+    "    res.loc[sample_date, 'UQER (# > 100%)'] = np.sum((explained_var / sec_var >= 1.))\n",
+    "    res.loc[sample_date, 'TF'] = tf_explained_per\n",
+    "    res.loc[sample_date, 'TF (std.)'] = np.std(tf_explained_var / sec_var)\n",
+    "    res.loc[sample_date, 'TF (# > 100%)'] = np.sum((tf_explained_var / sec_var >= 1.))\n",
+    "    res.loc[sample_date, 'avg. vol'] = avg_vol\n",
+    "    res.loc[sample_date, '# of Stk.'] = num_stks\n",
+    "    print(f\"{sample_date}: UQER: {explained_per:.2f}, {np.std(explained_var / sec_var):.2f}, {np.sum((explained_var / sec_var >= 1.)):0d} \\\n",
+    "        TF: {tf_explained_per:.2f}, {np.std(tf_explained_var / sec_var):.2f}, {np.sum((tf_explained_var / sec_var >= 1.)):0d}\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "res.to_excel('uqer_vs_tf.xlsx')"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "res[['UQER', 'TF']].rolling(window=30).mean().plot(secondary_y='avg. vol', figsize=(16, 8))"
+   ]
+  },
+  {
+   "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": "3.7.3"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}