main.py

# -*- coding: utf-8 -*-
import pymongo
import pandas as pd
import json
import datetime
import time
import numpy as np 


mytime = 1478635200
start = time.time()

dicoStates = {"Hawai":"HI","Alaska":"AK","Floride":"FL","New_Hampshire":"NH","Michigan":"MI","Vermont":"VT","Maine":"ME","Rhode_Island":"RI","New_York":"NY","Pennsylvanie":"PA","New_Jersey":"NJ","Delaware":"DE","Maryland":"MD","Virginie":"VA","Virginie_Occidentale":"WV","Ohio":"OH","Indiana":"IN","Illinois":"IL","Connecticut":"CT","Wisconsin":"WI","Caroline_du_Nord":"NC","District_de_Columbia":"DC","Massachusetts":"MA","Tennessee":"TN","Arkansas":"AR","Missouri":"MO","Georgie":"GA","Caroline_du_Sud":"SC","Kentucky":"KY","Alabama":"AL","Louisiane":"LS","Mississippi":"MS","Iowa":"IA","Minnesota":"MN","Oklahoma":"OK","Texas":"TX","Nouveau_Mexique":"NM","Kansas":"KS","Nebraska":"NE","Dakota_du_Sud":"SD","Dakota_du_Nord":"ND","Wyoming":"WY","Montana":"MT","Colorado":"CO","Idaho":"ID","Utah":"UT","Arizona":"AZ","Nevada":"NV","Oregon":"OR","Washington":"WA","Californie":"CA"}
c = pymongo.MongoClient('mongodb://172.31.31.100:27017,172.31.31.101:27017,172.31.31.102:27017/?replicaSet=rs0')

candidats = ["Trump", "Clinton", "Blanc"]
idx_section = 0

def load_data( p=.6): # proba de garder un état de sa couleur politique
    df = pd.read_csv("state.csv")
    probas_dem = []
    for gouv in df['gouv']:
        if gouv == 'Rép.':
            probas_dem.append(1 - p)
        elif gouv == 'Dém.':
            probas_dem.append(p)
        else:
            probas_dem.append(.5)
    df['proba_dem'] = probas_dem
    return df

def update_simulation(df, etat=None, gagnant=1):
    # a chaque résultat d'un état qui tombe on met à jour la grille des probas et on relance la simulation
    if etat != None:
        df.loc[np.argmax(df['State']==etat), 'proba_dem'] = gagnant # 1 si les démocrates ont pris l'état, 0 sinon
    # simulation monte carlo pour estimer la proba de victoire des démocrates
    B = 100
    dem = 0
    for i in range(0, B):
        if np.sum(df['nb_elector'][np.random.rand(51) < df['proba_dem']]) > 270:
            dem += 1

    result = dem /float(B)

    return result # proba que les démocrates gagnent les élections

df_proba = load_data(p=0.6)
list_proba_win = list()


while(1) :
    elapsed_time = time.time() - start
    #print(datetime.datetime.fromtimestamp(mytime+10*elapsed_time))
    result = list(c.elections.votes_2.find({"timestamp":{"$lte":datetime.datetime.fromtimestamp(mytime+30*elapsed_time)}}))
    df = pd.DataFrame(result)
    df["candidate"] = df["candidate"].map(lambda x: "Autre" if x not in candidats else x)
    df1 = df.groupby(["candidate", "state"]).sum()
    df1 = df1.reset_index()
    df1 = df1.pivot(index = "state", columns="candidate", values="voix")
    df1 = df1.reset_index()
    if "Blanc" not in list(df1.columns):
        df1["Blanc"] = 0

    df1  = df1.fillna(0) 
    df1["state"] = df1["state"].map(lambda x : dicoStates[x])
    df1["color"] = "#000000"
    df1["color"] = (np.argmax(df1[["Trump","Clinton"]].values, axis=1))
    df1["color"] = df1["color"].map(lambda x : "#FF0000" if x ==0 else "#3399FF")
    #df1 = df1.set_index("state")
    
    
    dfEmpty = pd.DataFrame(columns = ['state', "Trump", "Blanc", "Clinton", "Autre", "color"])
    
    dfEmpty["state"] = [elem for elem in dicoStates.values() if elem not in list(df1.state)]
    dfEmpty["Trump"] = 0
    dfEmpty["Blanc"] = 0
    dfEmpty["Clinton"] = 0
    dfEmpty["Autre"] = 0
    dfEmpty["color"] = "#FBF8EF"

    df1 = pd.concat([df1,dfEmpty])
    print(df1)
    dico = {}
    for col in df1.state.unique():
        dico2 = {}
        for item in ["Autre", "Blanc", "Clinton", "Trump", "color"]:
            dico2[item] = df1[df1["state"] == col][item].values[0]
        dico[col] = dico2

    print(dico)
    
    
    file = open("/var/www/html/donneesVotes2.json", "w")
    json.dump(dico,file)
    file.close()


# Section sur l'import des données aggrégés 
    data = json.load(open("/var/www/html/donneesVotes2.json"))
    base_elec = pd.read_csv("state.csv")

    result = pd.DataFrame(columns=["name","vote"])
    result = result.append({"name":"Clinton","vote":0}, ignore_index=True)
    result  = result.append({"name":"Trump","vote":0}, ignore_index=True)			

    proba = pd.DataFrame(columns=["name","percent"])
    proba = proba.append({"name":"Clinton","percent":0}, ignore_index=True)
    proba  = proba.append({"name":"Trump","percent":0}, ignore_index=True)


    for state in dicoStates.values():
# Les etats sont tous initialisés dans le json on doit boucler seulement sur les états dont les résultats sont >0
	try :
		vote_clinton = data[state]["Clinton"]
		vote_trump = data[state]["Trump"]

		vote_electeur = base_elec.loc[base_elec["State"]==state,"nb_elector"].values
                if vote_clinton > 0 and  vote_trump > 0:

			if vote_clinton > vote_trump :
				result.loc[result["name"]=="Clinton", "vote"] +=  vote_electeur[0]
                                list_proba_win.append( update_simulation(df_proba, state, gagnant=1)  )
			else :
				result.loc[result["name"]=="Trump", "vote"] += vote_electeur[0]
                                list_proba_win.append( update_simulation(df_proba, state, gagnant=0)  )

	except KeyError :
		continue


    print(result)
    idx_section +=1
    
    result.to_json("/var/www/html/jsonOne.json", orient="index") 
    print("Liste des probabilitées finales")
    print(list_proba_win)

    proba.loc[proba["name"]=="Clinton", "percent"] = list_proba_win[-1]
    proba.loc[proba["name"]=="Trump", "percent"] = (1 - list_proba_win[-1])

    print(proba) 
    proba.to_json("/var/www/html/jsonTwo.json", orient="index")
    print("Extraction numéro %s" % idx_section)


    if elapsed_time >= 3700 : break
    time.sleep(5)