1
0
mirror of https://github.com/gsi-upm/soil synced 2024-11-13 23:12:28 +00:00

Enterprise basic logic

This commit is contained in:
JesusMSM 2015-12-18 13:03:58 +01:00
parent a8e4e2efe6
commit 99090b1494
2 changed files with 34 additions and 16 deletions

View File

@ -45,8 +45,8 @@ def init():
##Users ##Users
tweet_probability_users = 0.44 tweet_probability_users = 0.44
tweet_relevant_probability = 0.25 tweet_relevant_probability = 0.25
tweet_probability_about = [0.25, 0.25] tweet_probability_about = [0, 0]
sentiment_about = [0, 0] sentiment_about = [0, 0] #Valores por defecto
##Enterprises ##Enterprises
tweet_probability_enterprises = [0.5, 0.5] tweet_probability_enterprises = [0.5, 0.5]

46
soil.py
View File

@ -7,6 +7,7 @@ import random
import numpy as np import numpy as np
import networkx as nx import networkx as nx
import settings import settings
import math
settings.init() # Loads all the data from settings settings.init() # Loads all the data from settings
@ -73,14 +74,9 @@ class BigMarketModel(BaseNetworkAgent):
def run(self): def run(self):
while True: while True:
if(self.id == 0 or self.id == 1): if(self.id < 2): # Empresa
# Empresa
print("He entrado a empresa")
self.enterpriseBehaviour() self.enterpriseBehaviour()
else: # Usuario
else:
# Usuario
print("He entrado a usuario")
self.userBehaviour() self.userBehaviour()
yield self.env.timeout(settings.timeout) yield self.env.timeout(settings.timeout)
@ -89,10 +85,19 @@ class BigMarketModel(BaseNetworkAgent):
def enterpriseBehaviour(self): def enterpriseBehaviour(self):
if random.random()< self.tweet_probability: #Twittea if random.random()< self.tweet_probability: #Twittea
aware_neighbors = self.get_neighboring_agents(state_id=2) aware_neighbors = self.get_neighboring_agents(state_id=2) #Nodos vecinos usuarios
for x in aware_neighbors: for x in aware_neighbors:
x.sentiment_about[0] += 0.01 #Aumenta para empresa 0 if random.uniform(0,10) < 5:
emotionStatus[x.id][self.env.now]=x.sentiment_about[0] x.sentiment_about[self.id] += 0.01 #Aumenta para empresa
else:
x.sentiment_about[self.id] -= 0.01 #Reduce para empresa
# Establecemos limites
if x.sentiment_about[self.id] > 1:
x.sentiment_about[self.id] = 1
if x.sentiment_about[self.id] < -1:
x.sentiment_about[self.id] = -1
#Guardamos estado para visualizacion
emotionStatus[x.id][self.env.now]=x.sentiment_about[self.id]
@ -101,6 +106,7 @@ class BigMarketModel(BaseNetworkAgent):
if random.random() < self.tweet_probability: #Twittea if random.random() < self.tweet_probability: #Twittea
if random.random() < self.tweet_relevant_probability: #Twittea algo relevante if random.random() < self.tweet_relevant_probability: #Twittea algo relevante
#Probabilidad de tweet para cada empresa #Probabilidad de tweet para cada empresa
for i in range(len(self.tweet_probability_about)): for i in range(len(self.tweet_probability_about)):
random_num = random.random() random_num = random.random()
@ -108,13 +114,23 @@ class BigMarketModel(BaseNetworkAgent):
#Se ha cumplido la condicion, evaluo los sentimientos hacia esa empresa #Se ha cumplido la condicion, evaluo los sentimientos hacia esa empresa
if self.sentiment_about[i] < 0: if self.sentiment_about[i] < 0:
#NEGATIVO #NEGATIVO
print("Sentimiento negativo") self.userTweets("negative",i)
elif self.sentiment_about[i] == 0: elif self.sentiment_about[i] == 0:
#NEUTRO #NEUTRO
print("Sentimiento neutro") pass
else: else:
#POSITIVO #POSITIVO
print("Sentimiento positivo") self.userTweets("positive",i)
def userTweets(self,sentiment,enterprise):
aware_neighbors = self.get_neighboring_agents(state_id=2) #Nodos vecinos usuarios
for x in aware_neighbors:
if sentiment == "positive":
x.sentiment_about[enterprise] +=0
elif sentiment == "negative":
x.sentiment_about[enterprise] -=0
else:
pass
@ -360,7 +376,9 @@ for x in range(0, settings.number_of_nodes):
emotionStatusAux=[] emotionStatusAux=[]
for tiempo in emotionStatus[x]: for tiempo in emotionStatus[x]:
if tiempo != 'id': if tiempo != 'id':
emotionStatusAux.append((emotionStatus[x][tiempo],tiempo,None)) prec = 2
output = math.floor(emotionStatus[x][tiempo] * (10 ** prec)) / (10 ** prec) #Para tener 2 decimales solo
emotionStatusAux.append((output,tiempo,None))
G.add_node(x, emotion= emotionStatusAux) G.add_node(x, emotion= emotionStatusAux)
#lista = nx.nodes(G) #lista = nx.nodes(G)