rating_training.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Predicción de calificaciones usando Random Forest.\n",
    "\n",
    "Problema supervisado. Cada observación corresponde a las variables asociadas a una compañía en una fecha particular. No se controla por fecha o industría.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1. Cargar librería requeridas"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "from rating_package.rating_prediction_functions import *"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2. Cargar archivos requeridos\n",
    "\n",
    "Para cargar los archivo de datos se debe tener en cuenta lo siguiente:\n",
    "\n",
    "* Todas las variables del archivo `research_data` que se desean incorporar al proceso de entrenamiento deben estar creadas en el archivo `features.csv`, y la llave corresponde a la columna `Key`. \n",
    "* Deben existir dos columnas para identificar las observaciones: `[Fecha, Ticker]`.\n",
    "* Debe existir variable `IssuerRating` que corresponde al output.\n",
    "* En caso de crear nuevas variables en el archivo `features.csv`, se deben evitar comas, tildes y otros símbolos."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Nombres de variables y llaves:\n",
    "feat_key = pd.read_csv('data/features.csv', sep=',', index_col = [\"Feature\"], encoding = \"latin1\")\n",
    "feat_key_pure = pd.read_csv('data/features_pure.csv', sep=',', index_col = [\"Feature\"], encoding = \"latin1\")\n",
    "\n",
    "# Encoder para calificaciones:\n",
    "le = pd.read_csv('data/lab_encoder.csv', sep=',', index_col = 0, encoding = \"latin1\")\n",
    "\n",
    "# Datos de entrenamiento:\n",
    "data_em = pd.read_csv('data/research_data_em.csv', sep=',', index_col = [\"Fecha\", 'Ticker'], encoding = \"latin1\")\n",
    "data_dm = pd.read_csv('data/research_data_dm.csv', sep=',', index_col = [\"Fecha\", 'Ticker'], encoding = \"latin1\")\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3. Proceso de entrenamiento.\n",
    "#### 3.1. Parámetros"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "remove_nan = True # Remover filas con datos faltantes.\n",
    "perc_train_size = 0.8 # Porcentaje de observaciones para entrenamiento.\n",
    "model_file = 'model/rating_random_forest_em.sav' # Modelo.\n",
    "model_file_pure = 'model/rating_random_forest_pure.sav' # Modelo.\n",
    "sov_encoder_file = 'model/sov_lab_encoder_em.sav' # Encoder de rating soberano.\n",
    "output_file = 'output/pred_test.csv' # Archivo de salida con prediciones.\n",
    "output_file_pure = 'output/pred_test_pure.csv' # Archivo de salida con prediciones.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 3.2. Entrenar modelo e imprimir resultados sobre muestra de testing\n",
    "Se publican los siguientes archivos: \n",
    "\n",
    "* El archivo en la ruta `output_file` contiene las observaciones usadas para testing, el rating real y el rating que predice el argoritmo.\n",
    "\n",
    "#### 3.2.1. Entrenamiento de modelo con datos de emergentes con Riesgo Soberano."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "c:\\users\\daniel.velasquez\\appdata\\local\\programs\\python\\python36\\lib\\site-packages\\sklearn\\model_selection\\_split.py:2026: FutureWarning: From version 0.21, test_size will always complement train_size unless both are specified.\n",
      "  FutureWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Muestra de entrenamiento: 1485\n",
      "Muestra de testing: 372\n",
      "\n",
      "Score sobre muestra de testing:\n",
      "0.8575268817204301\n",
      "\n",
      "\n",
      "Ranking:\n",
      "1. Rating Soberano País de Origen Deuda LP Moneda Extranjera (0.061807)\n",
      "2. Ingresos del Trimestre Anualizados / Total Activo (Promedio LTM) (0.055115)\n",
      "3. EBITDA del Trimestre Anualizado / Intereses (Promedio LTM) (0.046070)\n",
      "4. Ingresos del Trimestre Anualizados / Total Activo (0.045447)\n",
      "5. Activo Corto Plazo / Pasivo Corto Plazo (Promedio LTM) (0.044707)\n",
      "6. Deuda Neta / EBITDA del Trimestre Anualizado (Promedio LTM) (0.043551)\n",
      "7. Total Activo / Total Patrimonio (Promedio LTM) (0.042701)\n",
      "8. Total Activo / Total Patrimonio (0.042496)\n",
      "9. Margen EBITDA del Trimestre (Promedio LTM) (0.042187)\n",
      "10. Deuda de Corto Plazo / Deuda Total (Promedio LTM) (0.041952)\n",
      "11. Deuda Bruta / Activo (Promedio LTM) (0.041450)\n",
      "12. EBITDA del Trimestre Anualizado / Deuda Corto Plazo (Promedio LTM) (0.040153)\n",
      "13. Deuda Bruta / Activo (0.040079)\n",
      "14. Margen Neto del Trimestre (Promedio LTM) (0.036238)\n",
      "15. Utilidad Operacional del Trimestre / Activo Total (Promedio LTM) (0.035161)\n",
      "16. Activo Corto Plazo / Pasivo Corto Plazo (0.034907)\n",
      "17. Margen EBITDA del Trimestre (0.034032)\n",
      "18. EBITDA del Trimestre Anualizado / Intereses (0.033463)\n",
      "19. Deuda de Corto Plazo / Deuda Total (0.032496)\n",
      "20. Utilidad Neta del Trimestre Anualizada / Total Patrimonio (Promedio LTM) (0.031366)\n",
      "21. Deuda Neta / EBITDA del Trimestre Anualizado (0.030917)\n",
      "22. EBITDA del Trimestre Anualizado / Deuda Corto Plazo (0.029048)\n",
      "23. Crecimiento del EBITDA Últimos 12 Meses EBITDA (Promedio LTM) (0.028400)\n",
      "24. Utilidad Operacional del Trimestre / Activo Total (0.023527)\n",
      "25. Margen Neto del Trimestre (0.022586)\n",
      "26. Crecimiento del EBITDA Últimos 12 Meses EBITDA (0.020331)\n",
      "27. Utilidad Neta del Trimestre Anualizada / Total Patrimonio (0.019814)\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "c:\\users\\daniel.velasquez\\appdata\\local\\programs\\python\\python36\\lib\\site-packages\\sklearn\\preprocessing\\label.py:151: DeprecationWarning: The truth value of an empty array is ambiguous. Returning False, but in future this will result in an error. Use `array.size > 0` to check that an array is not empty.\n",
      "  if diff:\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "model_training(data_em, feat_key, le, remove_nan, perc_train_size, output_file, model_file, sov_encoder_file)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "#### 3.2.2. Entrenamiento de modelo con datos de desarrollados sin Riesgo Soberano."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "c:\\users\\daniel.velasquez\\appdata\\local\\programs\\python\\python36\\lib\\site-packages\\sklearn\\model_selection\\_split.py:2026: FutureWarning: From version 0.21, test_size will always complement train_size unless both are specified.\n",
      "  FutureWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Muestra de entrenamiento: 4535\n",
      "Muestra de testing: 1134\n",
      "\n",
      "Score sobre muestra de testing:\n",
      "0.8368606701940036\n",
      "\n",
      "\n",
      "Ranking:\n",
      "1. EBITDA del Trimestre Anualizado / Intereses (Promedio LTM) (0.059016)\n",
      "2. Activo Corto Plazo / Pasivo Corto Plazo (Promedio LTM) (0.052336)\n",
      "3. Ingresos del Trimestre Anualizados / Total Activo (Promedio LTM) (0.048452)\n",
      "4. Margen EBITDA del Trimestre (Promedio LTM) (0.047676)\n",
      "5. Deuda Bruta / Activo (Promedio LTM) (0.046285)\n",
      "6. Deuda Neta / EBITDA del Trimestre Anualizado (Promedio LTM) (0.044902)\n",
      "7. Total Activo / Total Patrimonio (Promedio LTM) (0.043136)\n",
      "8. Activo Corto Plazo / Pasivo Corto Plazo (0.042107)\n",
      "9. Margen Neto del Trimestre (Promedio LTM) (0.041517)\n",
      "10. Deuda Bruta / Activo (0.041484)\n",
      "11. Ingresos del Trimestre Anualizados / Total Activo (0.040567)\n",
      "12. Total Activo / Total Patrimonio (0.040545)\n",
      "13. Utilidad Operacional del Trimestre / Activo Total (Promedio LTM) (0.039862)\n",
      "14. EBITDA del Trimestre Anualizado / Intereses (0.039266)\n",
      "15. Deuda de Corto Plazo / Deuda Total (Promedio LTM) (0.039257)\n",
      "16. EBITDA del Trimestre Anualizado / Deuda Corto Plazo (Promedio LTM) (0.037448)\n",
      "17. Margen EBITDA del Trimestre (0.036367)\n",
      "18. Utilidad Neta del Trimestre Anualizada / Total Patrimonio (Promedio LTM) (0.035346)\n",
      "19. Deuda de Corto Plazo / Deuda Total (0.033182)\n",
      "20. Crecimiento del EBITDA Últimos 12 Meses EBITDA (Promedio LTM) (0.032610)\n",
      "21. Deuda Neta / EBITDA del Trimestre Anualizado (0.032434)\n",
      "22. EBITDA del Trimestre Anualizado / Deuda Corto Plazo (0.027935)\n",
      "23. Utilidad Operacional del Trimestre / Activo Total (0.026186)\n",
      "24. Margen Neto del Trimestre (0.025822)\n",
      "25. Crecimiento del EBITDA Últimos 12 Meses EBITDA (0.023688)\n",
      "26. Utilidad Neta del Trimestre Anualizada / Total Patrimonio (0.022573)\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "model_training(data_dm, feat_key_pure, le, remove_nan, perc_train_size, output_file_pure, model_file_pure, sov_encoder_file, n_estimators = 200)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "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
}