reports/report_2017-03-01.ipynb

   1 {
   2  "cells": [
   3   {
   4    "cell_type": "code",
   5    "execution_count": null,
   6    "metadata": {
   7     "collapsed": false
   8    },
   9    "outputs": [],
  10    "source": [
  11     "library(talweg)"
  12    ]
  13   },
  14   {
  15    "cell_type": "code",
  16    "execution_count": null,
  17    "metadata": {
  18     "collapsed": false
  19    },
  20    "outputs": [],
  21    "source": [
  22     "data = getData(ts_data=\"../data/pm10_mesures_H_loc.csv\", exo_data=\"../data/meteo_extra_noNAs.csv\",\n",
  23     "               input_tz = \"Europe/Paris\", working_tz=\"Europe/Paris\", predict_at=7)"
  24    ]
  25   },
  26   {
  27    "cell_type": "markdown",
  28    "metadata": {},
  29    "source": [
  30     "## Introduction\n",
  31     "\n",
  32     "J'ai fait quelques essais dans différentes configurations pour la méthode \"Neighbors\" (la seule dont on a parlé).<br>Il semble que le mieux soit\n",
  33     "\n",
  34     " * simtype=\"mix\" : on utilise les similarités endogènes et exogènes (fenêtre optimisée par VC)\n",
  35     " * same_season=FALSE : les indices pour la validation croisée ne tiennent pas compte des saisons\n",
  36     " * mix_strategy=\"mult\" : on multiplie les poids (au lieu d'en éteindre)\n",
  37     "\n",
  38     "J'ai systématiquement comparé à deux autres approches : la persistence et la moyenne de tous les futurs des jours similaires du passé ; à chaque fois sans prédiction du saut (sauf pour Neighbors : prédiction basée sur les poids calculés).\n",
  39     "\n",
  40     "Ensuite j'affiche les erreurs, quelques courbes prévues/mesurées, quelques filaments puis les histogrammes de quelques poids. Concernant les graphes de filaments, la moitié gauche du graphe correspond aux jours similaires au jour courant, tandis que la moitié droite affiche les lendemains : ce sont donc les voisinages tels qu'utilisés dans l'algorithme.\n",
  41     "\n",
  42     "<h2 style=\"color:blue;font-size:2em\">Pollution par chauffage</h2>"
  43    ]
  44   },
  45   {
  46    "cell_type": "code",
  47    "execution_count": null,
  48    "metadata": {
  49     "collapsed": false
  50    },
  51    "outputs": [],
  52    "source": [
  53     "p_ch_nn = getForecast(data, seq(as.Date(\"2015-01-18\"),as.Date(\"2015-01-24\"),\"days\"), Inf, 17,\n",
  54     "                   \"Neighbors\", \"Neighbors\", simtype=\"mix\", same_season=FALSE, mix_strategy=\"mult\")\n",
  55     "p_ch_pz = getForecast(data, seq(as.Date(\"2015-01-18\"),as.Date(\"2015-01-24\"),\"days\"), Inf, 17,\n",
  56     "                   \"Persistence\", \"Zero\")\n",
  57     "p_ch_az = getForecast(data, seq(as.Date(\"2015-01-18\"),as.Date(\"2015-01-24\"),\"days\"), Inf, 17,\n",
  58     "                   \"Average\", \"Zero\")"
  59    ]
  60   },
  61   {
  62    "cell_type": "code",
  63    "execution_count": null,
  64    "metadata": {
  65     "collapsed": false
  66    },
  67    "outputs": [],
  68    "source": [
  69     "e_ch_nn = getError(data, p_ch_nn, 17)\n",
  70     "e_ch_pz = getError(data, p_ch_pz, 17)\n",
  71     "e_ch_az = getError(data, p_ch_az, 17)\n",
  72     "options(repr.plot.width=9, repr.plot.height=6)\n",
  73     "plotError(list(e_ch_nn, e_ch_pz, e_ch_az), cols=c(1,2,colors()[258]))\n",
  74     "\n",
  75     "#Noir: neighbors, rouge: persistence, vert: moyenne"
  76    ]
  77   },
  78   {
  79    "cell_type": "markdown",
  80    "metadata": {},
  81    "source": [
  82     "La méthode Neighbors fait assez nettement mieux qu'une simple moyenne dans ce cas."
  83    ]
  84   },
  85   {
  86    "cell_type": "code",
  87    "execution_count": null,
  88    "metadata": {
  89     "collapsed": false
  90    },
  91    "outputs": [],
  92    "source": [
  93     "par(mfrow=c(1,2))\n",
  94     "options(repr.plot.width=9, repr.plot.height=4)\n",
  95     "plotPredReal(data, p_ch_nn, 3)\n",
  96     "plotPredReal(data, p_ch_nn, 4)\n",
  97     "\n",
  98     "#Bleu: prévue, noir: réalisée"
  99    ]
 100   },
 101   {
 102    "cell_type": "markdown",
 103    "metadata": {},
 104    "source": [
 105     "Prédictions d'autant plus lisses que le jour à prévoir est atypique (pollué)."
 106    ]
 107   },
 108   {
 109    "cell_type": "code",
 110    "execution_count": null,
 111    "metadata": {
 112     "collapsed": false
 113    },
 114    "outputs": [],
 115    "source": [
 116     "par(mfrow=c(1,2))\n",
 117     "plotFilaments(data, p_ch_nn$getIndexInData(3))\n",
 118     "plotFilaments(data, p_ch_nn$getIndexInData(4))"
 119    ]
 120   },
 121   {
 122    "cell_type": "markdown",
 123    "metadata": {},
 124    "source": [
 125     "Beaucoup de courbes similaires dans le cas peu pollué, très peu pour un jour pollué."
 126    ]
 127   },
 128   {
 129    "cell_type": "code",
 130    "execution_count": null,
 131    "metadata": {
 132     "collapsed": false
 133    },
 134    "outputs": [],
 135    "source": [
 136     "par(mfrow=c(1,3))\n",
 137     "plotSimils(p_ch_nn, 3)\n",
 138     "plotSimils(p_ch_nn, 4)\n",
 139     "plotSimils(p_ch_nn, 5)\n",
 140     "\n",
 141     "#Non pollué à gauche, pollué au milieu, autre pollué à droite"
 142    ]
 143   },
 144   {
 145    "cell_type": "markdown",
 146    "metadata": {},
 147    "source": [
 148     "La plupart des poids très proches de zéro ; pas pour le jour 5 : autre type de jour, cf. ci-dessous."
 149    ]
 150   },
 151   {
 152    "cell_type": "code",
 153    "execution_count": null,
 154    "metadata": {
 155     "collapsed": false
 156    },
 157    "outputs": [],
 158    "source": [
 159     "par(mfrow=c(1,2))\n",
 160     "plotPredReal(data, p_ch_nn, 5)\n",
 161     "plotFilaments(data, p_ch_nn$getIndexInData(5))"
 162    ]
 163   },
 164   {
 165    "cell_type": "markdown",
 166    "metadata": {},
 167    "source": [
 168     "<h2 style=\"color:blue;font-size:2em\">Pollution par épandage</h2>"
 169    ]
 170   },
 171   {
 172    "cell_type": "code",
 173    "execution_count": null,
 174    "metadata": {
 175     "collapsed": false
 176    },
 177    "outputs": [],
 178    "source": [
 179     "p_ep_nn = getForecast(data, seq(as.Date(\"2015-03-15\"),as.Date(\"2015-03-21\"),\"days\"), Inf, 17,\n",
 180     "                   \"Neighbors\", \"Neighbors\", simtype=\"mix\", same_season=FALSE, mix_strategy=\"mult\")\n",
 181     "p_ep_pz = getForecast(data, seq(as.Date(\"2015-03-15\"),as.Date(\"2015-03-21\"),\"days\"), Inf, 17,\n",
 182     "                   \"Persistence\", \"Zero\")\n",
 183     "p_ep_az = getForecast(data, seq(as.Date(\"2015-03-15\"),as.Date(\"2015-03-21\"),\"days\"), Inf, 17,\n",
 184     "                   \"Average\", \"Zero\")"
 185    ]
 186   },
 187   {
 188    "cell_type": "code",
 189    "execution_count": null,
 190    "metadata": {
 191     "collapsed": false
 192    },
 193    "outputs": [],
 194    "source": [
 195     "e_ep_nn = getError(data, p_ep_nn, 17)\n",
 196     "e_ep_pz = getError(data, p_ep_pz, 17)\n",
 197     "e_ep_az = getError(data, p_ep_az, 17)\n",
 198     "options(repr.plot.width=9, repr.plot.height=6)\n",
 199     "plotError(list(e_ep_nn, e_ep_pz, e_ep_az), cols=c(1,2,colors()[258]))\n",
 200     "\n",
 201     "#Noir: neighbors, rouge: persistence, vert: moyenne"
 202    ]
 203   },
 204   {
 205    "cell_type": "markdown",
 206    "metadata": {},
 207    "source": [
 208     "Cette fois les deux méthodes naïves font en moyenne moins d'erreurs que Neighbors. Prédiction trop difficile ?"
 209    ]
 210   },
 211   {
 212    "cell_type": "code",
 213    "execution_count": null,
 214    "metadata": {
 215     "collapsed": false
 216    },
 217    "outputs": [],
 218    "source": [
 219     "par(mfrow=c(1,2))\n",
 220     "options(repr.plot.width=9, repr.plot.height=4)\n",
 221     "plotPredReal(data, p_ep_nn, 4)\n",
 222     "plotPredReal(data, p_ep_nn, 6)"
 223    ]
 224   },
 225   {
 226    "cell_type": "markdown",
 227    "metadata": {},
 228    "source": [
 229     "À gauche un jour \"bien\" prévu, à droite le pic d'erreur (jour 6)."
 230    ]
 231   },
 232   {
 233    "cell_type": "code",
 234    "execution_count": null,
 235    "metadata": {
 236     "collapsed": false
 237    },
 238    "outputs": [],
 239    "source": [
 240     "par(mfrow=c(1,2))\n",
 241     "plotFilaments(data, p_ep_nn$getIndexInData(4))\n",
 242     "plotFilaments(data, p_ep_nn$getIndexInData(6))"
 243    ]
 244   },
 245   {
 246    "cell_type": "code",
 247    "execution_count": null,
 248    "metadata": {
 249     "collapsed": false
 250    },
 251    "outputs": [],
 252    "source": [
 253     "par(mfrow=c(1,2))\n",
 254     "plotSimils(p_ep_nn, 4)\n",
 255     "plotSimils(p_ep_nn, 6)"
 256    ]
 257   },
 258   {
 259    "cell_type": "markdown",
 260    "metadata": {},
 261    "source": [
 262     "Même observation concernant les poids : concentrés près de zéro pour les prédictions avec peu de voisins."
 263    ]
 264   },
 265   {
 266    "cell_type": "markdown",
 267    "metadata": {},
 268    "source": [
 269     "## Semaine non polluée"
 270    ]
 271   },
 272   {
 273    "cell_type": "code",
 274    "execution_count": null,
 275    "metadata": {
 276     "collapsed": false
 277    },
 278    "outputs": [],
 279    "source": [
 280     "p_np_nn = getForecast(data, seq(as.Date(\"2015-04-26\"),as.Date(\"2015-05-02\"),\"days\"), Inf, 17,\n",
 281     "                   \"Neighbors\", \"Neighbors\", simtype=\"mix\", same_season=FALSE, mix_strategy=\"mult\")\n",
 282     "p_np_pz = getForecast(data, seq(as.Date(\"2015-04-26\"),as.Date(\"2015-05-02\"),\"days\"), Inf, 17,\n",
 283     "                   \"Persistence\", \"Zero\")\n",
 284     "p_np_az = getForecast(data, seq(as.Date(\"2015-04-26\"),as.Date(\"2015-05-02\"),\"days\"), Inf, 17,\n",
 285     "                   \"Average\", \"Zero\")"
 286    ]
 287   },
 288   {
 289    "cell_type": "code",
 290    "execution_count": null,
 291    "metadata": {
 292     "collapsed": false
 293    },
 294    "outputs": [],
 295    "source": [
 296     "e_np_nn = getError(data, p_np_nn, 17)\n",
 297     "e_np_pz = getError(data, p_np_pz, 17)\n",
 298     "e_np_az = getError(data, p_np_az, 17)\n",
 299     "options(repr.plot.width=9, repr.plot.height=6)\n",
 300     "plotError(list(e_np_nn, e_np_pz, e_np_az), cols=c(1,2,colors()[258]))\n",
 301     "\n",
 302     "#Noir: neighbors, rouge: persistence, vert: moyenne"
 303    ]
 304   },
 305   {
 306    "cell_type": "markdown",
 307    "metadata": {},
 308    "source": [
 309     "Performances des méthodes \"Average\" et \"Neighbors\" comparables ; mauvais résultats pour la persistence."
 310    ]
 311   },
 312   {
 313    "cell_type": "code",
 314    "execution_count": null,
 315    "metadata": {
 316     "collapsed": false
 317    },
 318    "outputs": [],
 319    "source": [
 320     "par(mfrow=c(1,2))\n",
 321     "options(repr.plot.width=9, repr.plot.height=4)\n",
 322     "plotPredReal(data, p_np_nn, 3)\n",
 323     "plotPredReal(data, p_np_nn, 6)"
 324    ]
 325   },
 326   {
 327    "cell_type": "markdown",
 328    "metadata": {},
 329    "source": [
 330     "Les \"bonnes\" prédictions (à gauche) sont tout de même trop lissées."
 331    ]
 332   },
 333   {
 334    "cell_type": "code",
 335    "execution_count": null,
 336    "metadata": {
 337     "collapsed": false
 338    },
 339    "outputs": [],
 340    "source": [
 341     "par(mfrow=c(1,2))\n",
 342     "plotFilaments(data, p_np_nn$getIndexInData(3))\n",
 343     "plotFilaments(data, p_np_nn$getIndexInData(6))"
 344    ]
 345   },
 346   {
 347    "cell_type": "markdown",
 348    "metadata": {},
 349    "source": [
 350     "Jours \"typiques\", donc beaucoup de voisins."
 351    ]
 352   },
 353   {
 354    "cell_type": "code",
 355    "execution_count": null,
 356    "metadata": {
 357     "collapsed": false
 358    },
 359    "outputs": [],
 360    "source": [
 361     "par(mfrow=c(1,3))\n",
 362     "plotSimils(p_np_nn, 3)\n",
 363     "plotSimils(p_np_nn, 4)\n",
 364     "plotSimils(p_np_nn, 6)"
 365    ]
 366   },
 367   {
 368    "cell_type": "markdown",
 369    "metadata": {},
 370    "source": [
 371     "Répartition idéale des poids : quelques uns au-delà de 0.3-0.4, le reste très proche de zéro."
 372    ]
 373   },
 374   {
 375    "cell_type": "markdown",
 376    "metadata": {},
 377    "source": [
 378     "## Bilan\n",
 379     "\n",
 380     "Problème difficile : on ne fait guère mieux qu'une naïve moyenne des lendemains des jours similaires dans le passé.\n",
 381     "\n",
 382     "Comment améliorer la méthode ?"
 383    ]
 384   }
 385  ],
 386  "metadata": {
 387   "kernelspec": {
 388    "display_name": "R",
 389    "language": "R",
 390    "name": "ir"
 391   },
 392   "language_info": {
 393    "codemirror_mode": "r",
 394    "file_extension": ".r",
 395    "mimetype": "text/x-r-source",
 396    "name": "R",
 397    "pygments_lexer": "r",
 398    "version": "3.3.2"
 399   }
 400  },
 401  "nbformat": 4,
 402  "nbformat_minor": 2
 403 }