{"cells": [{"cell_type": "markdown", "metadata": {}, "source": ["# Elections et cartes \u00e9lectorales - \u00e9nonc\u00e9\n", "\n", "D'apr\u00e8s wikip\u00e9dia, le [Gerrymandering](https://fr.wikipedia.org/wiki/Gerrymandering) est un terme politique nord-am\u00e9ricain pour d\u00e9signer le d\u00e9coupage des circonscriptions \u00e9lectorales ayant pour objectif de donner l\u2019avantage \u00e0 un parti, un candidat, ou un groupe donn\u00e9. Et c'est ce que nous allons faire dans cette s\u00e9ance. C'est un probl\u00e8me tout-\u00e0-fait d'actualit\u00e9 : [Primaire de la droite : 10 228 bureaux de vote strat\u00e9giquement r\u00e9partis](http://www.lemonde.fr/election-presidentielle-2017/article/2016/09/30/primaire-de-la-droite-10-228-bureaux-de-vote-strategiquement-repartis_5005831_4854003.html)."]}, {"cell_type": "code", "execution_count": 1, "metadata": {"ExecuteTime": {"end_time": "2016-10-30T09:50:46.959642", "start_time": "2016-10-30T09:50:46.230655"}}, "outputs": [], "source": ["%matplotlib inline"]}, {"cell_type": "code", "execution_count": 2, "metadata": {"ExecuteTime": {"end_time": "2016-10-30T09:50:46.977649", "start_time": "2016-10-30T09:50:46.962642"}}, "outputs": [{"data": {"text/html": ["
run previous cell, wait for 2 seconds
\n", ""], "text/plain": [""]}, "execution_count": 3, "metadata": {}, "output_type": "execute_result"}], "source": ["from jyquickhelper import add_notebook_menu\n", "add_notebook_menu()"]}, {"cell_type": "markdown", "metadata": {}, "source": ["## Donn\u00e9es\n", "\n", "Les donn\u00e9es sont de plusieurs types et sont regroup\u00e9es en trois fichiers :\n", "\n", "* [R\u00e9sultat des \u00e9lections l\u00e9gislatives fran\u00e7aises de 2012 au niveau bureau de vote](https://www.data.gouv.fr/fr/datasets/resultat-des-elections-legislatives-francaises-de-2012-au-niveau-bureau-de-vote-nd/)\n", "* [Countours des circonscriptions des l\u00e9gislatives](https://www.data.gouv.fr/fr/datasets/countours-des-circonscriptions-des-legislatives-nd/)\n", "* [Localisation des buraux de votes](http://www.xavierdupre.fr/enseignement/complements/bureauxvotegeo.zip)\n", "* [Localisation des villes](https://github.com/sdpython/actuariat_python/raw/master/src/actuariat_python/data/data_elections/villes_geo.zip)\n", "* [Localisation des bureaux de vote avec Cartelec](http://www.cartelec.net/?page_id=3609), cette base requiert la conversion des coordonn\u00e9es (lire ce [notebook](http://www.xavierdupre.fr/app/ensae_teaching_cs/helpsphinx/notebooks/td1a_cenonce_session_12.html#cartographie))\n", "\n", "Il est conseill\u00e9 de t\u00e9l\u00e9charger directement ces donn\u00e9es. Les paragraphes suivants expliquent comment ceux-ci ont \u00e9t\u00e9 r\u00e9cup\u00e9r\u00e9s ou construit. Il n'est pas imm\u00e9diat d'obtenir la localisation des bureaux de vote. Celle-ci n'est d'ailleurs pas compl\u00e8te."]}, {"cell_type": "markdown", "metadata": {}, "source": ["### R\u00e9sultats des \u00e9lections l\u00e9gislatives"]}, {"cell_type": "code", "execution_count": 3, "metadata": {"ExecuteTime": {"end_time": "2016-10-30T09:50:52.596577", "start_time": "2016-10-30T09:50:46.980646"}}, "outputs": [], "source": ["from actuariat_python.data import elections_legislatives_bureau_vote\n", "tour = elections_legislatives_bureau_vote(source='xd')"]}, {"cell_type": "code", "execution_count": 4, "metadata": {"ExecuteTime": {"end_time": "2016-10-30T09:50:52.771571", "start_time": "2016-10-30T09:50:52.602572"}}, "outputs": [{"data": {"text/html": ["
\n", "\n", "
\n", " \n", "
\n", "
\n", "
N\u00b0 tour
\n", "
Code d\u00e9partement
\n", "
Code de la commune
\n", "
Nom de la commune
\n", "
N\u00b0 de circonscription Lg
\n", "
N\u00b0 de canton
\n", "
N\u00b0 de bureau de vote
\n", "
Inscrits
\n", "
Votants
\n", "
Exprim\u00e9s
\n", "
N\u00b0 de d\u00e9p\u00f4t du candidat
\n", "
Nom du candidat
\n", "
Pr\u00e9nom du candidat
\n", "
Code nuance du candidat
\n", "
Nombre de voix du candidat
\n", "
\n", " \n", " \n", "
\n", "
3858
\n", "
2
\n", "
01
\n", "
16
\n", "
Arbigny
\n", "
1
\n", "
26
\n", "
0001
\n", "
309
\n", "
146
\n", "
144
\n", "
32
\n", "
BRETON
\n", "
Xavier
\n", "
UMP
\n", "
87
\n", "
\n", "
\n", "
3859
\n", "
2
\n", "
01
\n", "
16
\n", "
Arbigny
\n", "
1
\n", "
26
\n", "
0001
\n", "
309
\n", "
146
\n", "
144
\n", "
33
\n", "
DEBAT
\n", "
Jean-Fran\u00e7ois
\n", "
SOC
\n", "
57
\n", "
\n", "
\n", "
3871
\n", "
2
\n", "
01
\n", "
24
\n", "
Attignat
\n", "
1
\n", "
21
\n", "
0001
\n", "
746
\n", "
425
\n", "
411
\n", "
32
\n", "
BRETON
\n", "
Xavier
\n", "
UMP
\n", "
233
\n", "
\n", "
\n", "
3872
\n", "
2
\n", "
01
\n", "
24
\n", "
Attignat
\n", "
1
\n", "
21
\n", "
0001
\n", "
746
\n", "
425
\n", "
411
\n", "
33
\n", "
DEBAT
\n", "
Jean-Fran\u00e7ois
\n", "
SOC
\n", "
178
\n", "
\n", "
\n", "
3873
\n", "
2
\n", "
01
\n", "
24
\n", "
Attignat
\n", "
1
\n", "
21
\n", "
0002
\n", "
1311
\n", "
801
\n", "
792
\n", "
32
\n", "
BRETON
\n", "
Xavier
\n", "
UMP
\n", "
446
\n", "
\n", " \n", "
\n", "
"], "text/plain": [" N\u00b0 tour Code d\u00e9partement Code de la commune Nom de la commune \\\n", "3858 2 01 16 Arbigny \n", "3859 2 01 16 Arbigny \n", "3871 2 01 24 Attignat \n", "3872 2 01 24 Attignat \n", "3873 2 01 24 Attignat \n", "\n", " N\u00b0 de circonscription Lg N\u00b0 de canton N\u00b0 de bureau de vote Inscrits \\\n", "3858 1 26 0001 309 \n", "3859 1 26 0001 309 \n", "3871 1 21 0001 746 \n", "3872 1 21 0001 746 \n", "3873 1 21 0002 1311 \n", "\n", " Votants Exprim\u00e9s N\u00b0 de d\u00e9p\u00f4t du candidat Nom du candidat \\\n", "3858 146 144 32 BRETON \n", "3859 146 144 33 DEBAT \n", "3871 425 411 32 BRETON \n", "3872 425 411 33 DEBAT \n", "3873 801 792 32 BRETON \n", "\n", " Pr\u00e9nom du candidat Code nuance du candidat Nombre de voix du candidat \n", "3858 Xavier UMP 87 \n", "3859 Jean-Fran\u00e7ois SOC 57 \n", "3871 Xavier UMP 233 \n", "3872 Jean-Fran\u00e7ois SOC 178 \n", "3873 Xavier UMP 446 "]}, "execution_count": 5, "metadata": {}, "output_type": "execute_result"}], "source": ["tour[\"T2\"].sort_values([\"Code d\u00e9partement\", \"N\u00b0 de circonscription Lg\"]).head()"]}, {"cell_type": "markdown", "metadata": {}, "source": ["### G\u00e9olocalisation des circonscription"]}, {"cell_type": "code", "execution_count": 5, "metadata": {"ExecuteTime": {"end_time": "2016-10-30T09:50:52.993566", "start_time": "2016-10-30T09:50:52.778575"}}, "outputs": [], "source": ["from actuariat_python.data import elections_legislatives_circonscription_geo\n", "geo = elections_legislatives_circonscription_geo()"]}, {"cell_type": "code", "execution_count": 6, "metadata": {"ExecuteTime": {"end_time": "2016-10-30T09:50:53.095565", "start_time": "2016-10-30T09:50:52.999565"}}, "outputs": [{"data": {"text/html": ["
groupe scolaire saint-exup\u00e9ry - salle de jeux,...
\n", "
1000
\n", "
rue antoine de saint exup\u00e9ry 01000 bourg-en-br...
\n", "
46.210030
\n", "
5.233330
\n", "
Rue Antoine de Saint-Exup\u00e9ry, Bourg-en-Bresse,...
\n", "
\n", "
\n", "
3
\n", "
11 avenue de l\u2019\u00e9galit\u00e9
\n", "
bourg-en-bresse
\n", "
4
\n", "
charles perrault - \u00e9cole primaire
\n", "
1000
\n", "
11 avenue de l\u2019\u00e9galit\u00e9 01000 bourg-en-bresse
\n", "
46.214848
\n", "
5.231941
\n", "
11, Avenue de l'\u00c9galit\u00e9, Saint-Georges, La Gla...
\n", "
\n", "
\n", "
4
\n", "
11 avenue de l\u2019\u00e9galit\u00e9
\n", "
bourg-en-bresse
\n", "
5
\n", "
charles perrault - \u00e9cole primaire
\n", "
1000
\n", "
11 avenue de l\u2019\u00e9galit\u00e9 01000 bourg-en-bresse
\n", "
46.214848
\n", "
5.231941
\n", "
11, Avenue de l'\u00c9galit\u00e9, Saint-Georges, La Gla...
\n", "
\n", " \n", "
\n", "
"], "text/plain": [" address city n \\\n", "0 cours verdun bourg 1 \n", "1 cours verdun bourg 2 \n", "2 rue antoine de saint exup\u00e9ry bourg-en-bresse 3 \n", "3 11 avenue de l\u2019\u00e9galit\u00e9 bourg-en-bresse 4 \n", "4 11 avenue de l\u2019\u00e9galit\u00e9 bourg-en-bresse 5 \n", "\n", " place zip \\\n", "0 salle des f\u00eates 1000 \n", "1 salle des f\u00eates 1000 \n", "2 groupe scolaire saint-exup\u00e9ry - salle de jeux,... 1000 \n", "3 charles perrault - \u00e9cole primaire 1000 \n", "4 charles perrault - \u00e9cole primaire 1000 \n", "\n", " full_address latitude longitude \\\n", "0 cours verdun 01000 bourg 46.206605 5.228364 \n", "1 cours verdun 01000 bourg 46.206605 5.228364 \n", "2 rue antoine de saint exup\u00e9ry 01000 bourg-en-br... 46.210030 5.233330 \n", "3 11 avenue de l\u2019\u00e9galit\u00e9 01000 bourg-en-bresse 46.214848 5.231941 \n", "4 11 avenue de l\u2019\u00e9galit\u00e9 01000 bourg-en-bresse 46.214848 5.231941 \n", "\n", " geo_address \n", "0 Cours de Verdun, Le Peloux, Les Vennes, Bourg-... \n", "1 Cours de Verdun, Le Peloux, Les Vennes, Bourg-... \n", "2 Rue Antoine de Saint-Exup\u00e9ry, Bourg-en-Bresse,... \n", "3 11, Avenue de l'\u00c9galit\u00e9, Saint-Georges, La Gla... \n", "4 11, Avenue de l'\u00c9galit\u00e9, Saint-Georges, La Gla... "]}, "execution_count": 11, "metadata": {}, "output_type": "execute_result"}], "source": ["bureau_geo.head()"]}, {"cell_type": "markdown", "metadata": {}, "source": ["**Ce qui suit explique la fa\u00e7on dont j'ai constuit cette table.**\n", "\n", "Les bureaux sont assez stables d'une \u00e9lection \u00e0 l'autre et cela ne devrait pas trop poser de probl\u00e8mes si on m\u00e9lange les donn\u00e9es. En revanche, ces donn\u00e9es sont assez difficiles \u00e0 obtenir. [open.data.fr](https://www.data.gouv.fr/fr/datasets/?q=bureaux+de+vote&page=1) propose ces donn\u00e9es mais il faut r\u00e9cup\u00e9rer chaque ville ou chaque r\u00e9gion s\u00e9par\u00e9ment sans garantie de r\u00e9ussir \u00e0 couvrir tout le territoire. Le site [NosDonnes.fr](http://www.nosdonnees.fr/dataset/golocalisation-des-bureaux-de-votes) recense bien toutes ces informations mais il n'est pas possible - au moment o\u00f9 j'\u00e9cris ces lignes - de r\u00e9cup\u00e9rer ces donn\u00e9es sous la forme d'un fichier plat. De plus, certaines r\u00e9gions ne sont disponibles que sous forme de scan d'impressions papier. C'est en lisant l'article [Comment red\u00e9coupe-t-on la carte \u00e9lectorale?](http://www.slate.fr/story/3945/comment-redecouper-t-la-carte-electorale) que je suis tomb\u00e9 finalement sur la base constitu\u00e9e pour r\u00e9diger l'article [Etude sur le red\u00e9coupage \u00e9lectoral : \u00e9valuer le poids politique de la r\u00e9forme](https://www.regardscitoyens.org/etude-sur-le-redecoupage-electoral-poids-politique-reforme/). Les donn\u00e9es sont accessibles sur [RegardsCitoyens.fr](https://www.regardscitoyens.org/publication/#donnees) et en cherchant bien, on trouve le r\u00e9pertoire [red\u00e9coupage](https://www.regardscitoyens.org/telechargement/redecoupage/) et le fichier [2014041401_resultats.csv.zip](https://www.regardscitoyens.org/telechargement/redecoupage/2014041401_resultats.csv.zip). Cependant, la g\u00e9olocalisation des bureaux n'est pas souvent renseign\u00e9e. On peut se contenter des fichiers obtenus pour quelques zones seulement : [Paris](https://www.data.gouv.fr/fr/datasets/liste-des-bureaux-de-vote-prs/), [Gironde](https://www.data.gouv.fr/fr/datasets/geo-localisation-des-bureaux-de-vote-du-departement-de-la-gironde-rdl/), [Montpellier](https://www.data.gouv.fr/fr/datasets/bureaux-de-vote-mtn/), [Marseille](https://www.data.gouv.fr/fr/datasets/bureaux-de-vote-de-marseille-op/), [Saint-Malo](https://www.data.gouv.fr/fr/datasets/emplacement-des-bureaux-de-vote/), [Nogent-Sur-Marne](https://www.data.gouv.fr/fr/datasets/bureaux-de-vote-de-la-ville-de-nogent-sur-marne/), [Haut-de-Seine](https://www.data.gouv.fr/fr/datasets/bureaux-de-vote-7/), [Toulouse](https://www.data.gouv.fr/fr/datasets/bureaux-de-vote-2012/), [Grand-Poitiers](https://www.data.gouv.fr/fr/datasets/citoyennete-bureaux-de-vote-grand-poitiers-donnees-de-reference/), [Calvados](https://www.data.gouv.fr/fr/datasets/bureaux-centralisateurs-des-circonscriptions-cantonales-dans-le-calvados/). Cette approche risque d'\u00eatre fastidieuse dans la mesure o\u00f9 les formats pour chaque ville ont de grande chance d'\u00eatre diff\u00e9rent. \n", "\n", "La meilleure option est peut-\u00eatre de scrapper le site [bureaudevote.fr](http://bureaudevote.fr/bureaudevote00.htm) - qui fonctionne bien quand il n'est pas en maintenance - en esp\u00e9rant que les num\u00e9ros des bureaux de votes correspondent. Le site propose seulement les adresses. Il faudra utiliser une API de geocoding. Quelque soit la source, on supposera alors que tous les bureaux de vote associ\u00e9s au m\u00eame emplacement feront n\u00e9cessairement partie de la m\u00eame circonscription. Bref, **l'open data passe aussi par une certaine standardisation !**\n", "\n", "Certains d\u00e9partements ne sont pas renseign\u00e9s, la Dr\u00f4me par exemple. Il faut aller sur d'autres sites comme [linternaute.com](http://www.linternaute.com/ville/valence/ville-26362/bureaux-vote#ville-26362-1), acc\u00e9der \u00e0 la [carte](http://www.linternaute.com/ville/p-bureau-vote#trouver-bureau-vote), mais il faudra un peu plus de temps pour r\u00e9cup\u00e9rer toutes ces informations. Le code propos\u00e9 ci-dessus r\u00e9cup\u00e8re les coordonn\u00e9es des bureaux de vote. Cela prend plusieurs heures et il faut relancer le processus quand il s'arr\u00eate pour une erreur de r\u00e9seau. Il est conseill\u00e9 de t\u00e9l\u00e9charger le r\u00e9sultat. Le g\u00e9ocodeur d'[OpenStreetMap](http://openstreetmap.fr/) n'est pas de tr\u00e8s bonne qualit\u00e9 sur la France. Il ne retourne rien dans plus d'un tiers des cas. On peut compl\u00e9ter avec l'[API de Bing Maps](https://www.microsoft.com/maps/choose-your-bing-maps-API.aspx)."]}, {"cell_type": "code", "execution_count": 11, "metadata": {"ExecuteTime": {"end_time": "2016-10-30T09:51:03.872343", "start_time": "2016-10-30T09:50:53.661561"}}, "outputs": [], "source": ["from actuariat_python.data import elections_vote_place_address\n", "bureau = elections_vote_place_address(hide_warning=True)"]}, {"cell_type": "code", "execution_count": 12, "metadata": {"ExecuteTime": {"end_time": "2016-10-30T09:51:03.923346", "start_time": "2016-10-30T09:51:03.875344"}}, "outputs": [{"data": {"text/html": ["
\n", "\n", "
\n", " \n", "
\n", "
\n", "
address
\n", "
city
\n", "
n
\n", "
place
\n", "
zip
\n", "
\n", " \n", " \n", "
\n", "
0
\n", "
cours verdun
\n", "
bourg
\n", "
1
\n", "
salle des f\u00eates
\n", "
01000
\n", "
\n", "
\n", "
1
\n", "
cours verdun
\n", "
bourg
\n", "
2
\n", "
salle des f\u00eates
\n", "
01000
\n", "
\n", "
\n", "
2
\n", "
rue antoine de saint exup\u00e9ry
\n", "
bourg-en-bresse
\n", "
3
\n", "
groupe scolaire saint-exup\u00e9ry - salle de jeux,...
\n", "
01000
\n", "
\n", "
\n", "
3
\n", "
11 avenue de l\u2019\u00e9galit\u00e9
\n", "
bourg-en-bresse
\n", "
4
\n", "
charles perrault - \u00e9cole primaire
\n", "
01000
\n", "
\n", "
\n", "
4
\n", "
11 avenue de l\u2019\u00e9galit\u00e9
\n", "
bourg-en-bresse
\n", "
5
\n", "
charles perrault - \u00e9cole primaire
\n", "
01000
\n", "
\n", " \n", "
\n", "
"], "text/plain": [" address city n \\\n", "0 cours verdun bourg 1 \n", "1 cours verdun bourg 2 \n", "2 rue antoine de saint exup\u00e9ry bourg-en-bresse 3 \n", "3 11 avenue de l\u2019\u00e9galit\u00e9 bourg-en-bresse 4 \n", "4 11 avenue de l\u2019\u00e9galit\u00e9 bourg-en-bresse 5 \n", "\n", " place zip \n", "0 salle des f\u00eates 01000 \n", "1 salle des f\u00eates 01000 \n", "2 groupe scolaire saint-exup\u00e9ry - salle de jeux,... 01000 \n", "3 charles perrault - \u00e9cole primaire 01000 \n", "4 charles perrault - \u00e9cole primaire 01000 "]}, "execution_count": 13, "metadata": {}, "output_type": "execute_result"}], "source": ["bureau.head()"]}, {"cell_type": "markdown", "metadata": {}, "source": ["On r\u00e9cup\u00e8re une cl\u00e9 pour utiliser l'[API de Bing Maps](https://www.microsoft.com/maps/choose-your-bing-maps-API.aspx) avec le module [keyring](http://pythonhosted.org/keyring/). Pour stocker son mot de passe sur la machine, il suffit d'\u00e9crire :\n", "\n", "```\n", "import keyring\n", "keyring.get_password(\"bing\", \"actuariat_python,key\")\n", "```"]}, {"cell_type": "code", "execution_count": 13, "metadata": {"ExecuteTime": {"end_time": "2016-10-30T09:51:04.253339", "start_time": "2016-10-30T09:51:03.929345"}}, "outputs": [{"data": {"text/plain": ["1"]}, "execution_count": 14, "metadata": {}, "output_type": "execute_result"}], "source": ["import keyring, os\n", "bing_key = keyring.get_password(\"bing\", \"actuariat_python,key\")\n", "coders = [\"Nominatim\"]\n", "if bing_key:\n", " # si la cl\u00e9 a \u00e9t\u00e9 trouv\u00e9e\n", " coders.append((\"bing\", bing_key))\n", "len(coders)"]}, {"cell_type": "code", "execution_count": 14, "metadata": {"ExecuteTime": {"end_time": "2016-10-30T09:51:04.328340", "start_time": "2016-10-30T09:51:04.256341"}}, "outputs": [{"name": "stdout", "output_type": "stream", "text": ["Les donn\u00e9es ont d\u00e9j\u00e0 \u00e9t\u00e9 geocod\u00e9es.\n"]}], "source": ["import os\n", "if not os.path.exists(\"bureauxvotegeo.zip\"):\n", " from actuariat_python.data import geocode\n", " from pyquickhelper.loghelper import fLOG\n", " fLOG(OutputPrint=True)\n", " bureau_geo = geocode(bureau, fLOG=fLOG, index=False, encoding=\"utf-8\",\n", " exc=False, save_every=\"bureau.dump.txt\", sep=\"\\t\", every=100,\n", " coders=coders)\n", "else:\n", " print(\"Les donn\u00e9es ont d\u00e9j\u00e0 \u00e9t\u00e9 geocod\u00e9es.\")"]}, {"cell_type": "markdown", "metadata": {}, "source": ["On regarde les valeurs manquantes."]}, {"cell_type": "code", "execution_count": 15, "metadata": {"ExecuteTime": {"end_time": "2016-10-30T09:51:17.653166", "start_time": "2016-10-30T09:51:04.333340"}}, "outputs": [{"data": {"image/png": "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\n", "text/plain": ["