コード例 #1
0
def Plot_Social_Welfare(task_n, provider_n, max_value, max_alpha, max_deadline, max_task_size, max_mu, max_provider_bid, max_provider_skill, task_columns, provider_columns):
  
  tasks = TaskCreator(task_n, max_value, max_alpha, max_deadline, max_task_size)
  providers = ProviderCreator(provider_n, max_mu, max_provider_bid, max_provider_skill)
  
  pro_exp_social = []
  exp_social = []
  pro_real_social = []
  real_social = []
  
  for i in range(1,11):
    capacity = 20*i
    
    auctioneer = Platform(capacity)

  #platform에서 winners를 선택한다. 
    W_requesters, req_threshold = auctioneer.WinningRequesterSelection(tasks)
    W_providers, pro_threshold = auctioneer.WinningProviderSelection(providers)

  #혹시 모르니깐 requesters와 providers의 숫자를 맞춘다.
    W_requesters, W_providers = auctioneer.Trimming(W_requesters, W_providers)

  #requesters와 providers의 preference ordering을 만든다.
    requester_preference_ordering = auctioneer.ordering_matrix(W_providers,provider_columns)
    provider_preference_ordering = auctioneer.ordering_matrix(W_requesters,task_columns)

    b_rank, b_mean_rank = benchmark_satisfaction_level(requester_preference_ordering)
    b_pdf, b_cdf = bench_distribution(b_rank)

    match = auctioneer.SMA(requester_preference_ordering, provider_preference_ordering)
    #In match, requester: keys, providers: values

    #expected social welfare comparison
    social_welfare = SocialWelfare(W_requesters, W_providers, 0.01, match)
    benchmark = SystemExpectedSocialwelfare(W_requesters, W_providers, 0.01)
    
    pro_exp_social.append(social_welfare)
    exp_social.append(benchmark)
    
    rank, mean_rank = auctioneer.satisfaction_level(match, requester_preference_ordering)
    pdf, cdf = auctioneer.satisfaction_distribution(rank)
    
    #real social welfare
    welfare, t_sub = Proposed_SocialWelfare(W_requesters, W_providers, 0.01, match)
    bench_welfare = Existing_SocialWelfare(W_requesters, W_providers, 0.01)
    
    pro_real_social.append(welfare)
    real_social.append(bench_welfare)
  
  x_axis = np.array(range(1,11))*20
  bar_width = 10
  fig, ax = plt.subplots(nrows = 1, ncols = 2)
  ax[0].bar(x_axis, pro_exp_social, width = bar_width, label = 'proposed exp.SW', color = 'b')
  ax[0].bar(x_axis+bar_width, exp_social, width = bar_width, label = 'benchmark exp.SW', color = 'r')
  ax[0].legend(loc = 'best')
  
  ax[1].bar(x_axis, pro_real_social, width = bar_width, label = 'proposed real.SW', color = 'b')
  ax[1].bar(x_axis+bar_width, real_social, width = bar_width, label = 'benchmark real.SW', color = 'r')
  ax[1].legend(loc = 'best')
  plt.show()  
コード例 #2
0
def preference_cdf(time_unit, task_n, provider_n, capacity, max_value, max_deadline, max_alpha, max_task_size, max_provider_bid, max_provider_skill, max_mu, task_columns, provider_columns, iteration, output): 

  r_distribution1 = []
  r_distribution2 = []
  r_distribution3 = []

  r_cumulative1 = []
  r_cumulative2 = []
  r_cumulative3 = []

  p_distribution1 = []
  p_distribution2 = []
  p_distribution3 = []

  p_cumulative1 = []
  p_cumulative2 = []
  p_cumulative3 = []

  r_mean1=[]
  r_mean2=[]
  r_mean3=[]

  p_mean1=[]
  p_mean2=[]
  p_mean3=[]

  for _ in range(iteration):
  
    tasks = TaskCreator(task_n, max_value, max_alpha, max_deadline, max_task_size)
    providers = ProviderCreator(provider_n, max_mu, max_provider_bid, max_provider_skill)

    auctioneer = Platform(capacity)
  
#without consideration to alpha and mu '''winner selection'''
    W_requesters, req_threshold = auctioneer.WinningRequesterSelection(tasks)
    W_providers, pro_threshold = auctioneer.WinningProviderSelection(providers)

#with consideration to alpha and mu
    New_W_requesters, New_req_threshold = auctioneer.New_WinningRequesterSelection(tasks)
    New_W_providers, New_pro_threshold = auctioneer.New_WinningProviderSelection(providers)

#trimming process
    W_requesters, W_providers = auctioneer.Trimming(W_requesters, W_providers)
    New_W_requesters, New_W_providers = auctioneer.Trimming(New_W_requesters, New_W_providers)



#1. preference ordering

#satisfaction level without consideration of preference, alpha and mu: 1
    requester_preference_ordering = auctioneer.ordering_matrix(W_providers,provider_columns)
    provider_preference_ordering = auctioneer.ordering_matrix(W_requesters,task_columns)

    r_rank, r_mean_rank = benchmark_satisfaction_level(requester_preference_ordering)
    p_rank, p_mean_rank = benchmark_satisfaction_level(provider_preference_ordering)
  
    r_pdf1, r_cdf1 = bench_distribution(r_rank)
    p_pdf1, p_cdf1 = bench_distribution(p_rank)
  
    r_distribution1.append(r_pdf1)
    r_cumulative1.append(r_cdf1)
  
    p_distribution1.append(p_pdf1)
    p_cumulative1.append(p_cdf1)
  
    r_mean1.append(r_mean_rank)
    p_mean1.append(p_mean_rank)
  
#satisfaction level without consideration of preference but with alpha, mu: 2
    New_requester_preference_ordering = auctioneer.ordering_matrix(New_W_providers,provider_columns)
    New_provider_preference_ordering = auctioneer.ordering_matrix(New_W_requesters,task_columns)

    New_r_rank, New_r_mean_rank = benchmark_satisfaction_level(New_requester_preference_ordering)
    New_p_rank, New_p_mean_rank = benchmark_satisfaction_level(New_provider_preference_ordering)
  
    r_pdf2, r_cdf2 = bench_distribution(New_r_rank)
    p_pdf2, p_cdf2 = bench_distribution(New_p_rank)

    r_distribution2.append(r_pdf2)
    p_distribution2.append(p_pdf2)
  
    r_cumulative2.append(r_cdf2)
    p_cumulative2.append(p_cdf2)
  
    r_mean2.append(New_r_mean_rank)
    p_mean2.append(New_p_mean_rank)
  
#Run Stable Marriage Algorithm: In match, requester: keys, providers: values
    match = auctioneer.SMA(New_requester_preference_ordering, New_provider_preference_ordering)



#satisfaction with consideration to preference, alpha, mu:3 
#requesters
    r_rank3, r_mean_rank3 = auctioneer.satisfaction_level(match, New_requester_preference_ordering)
  
#providers
    p_rank3, p_mean_rank3 = auctioneer.satisfaction_level(match, New_provider_preference_ordering)

    r_pdf3, r_cdf3 = auctioneer.satisfaction_distribution(r_rank3)
    p_pdf3, p_cdf3 = auctioneer.satisfaction_distribution(p_rank3)
  
    r_distribution3.append(r_pdf3)
    p_distribution3.append(p_pdf3)
  
    r_cumulative3.append(r_cdf3)
    p_cumulative3.append(p_cdf3)
  
    r_mean3.append(r_mean_rank3)
    p_mean3.append(p_mean_rank3)

  r_distribution1 = np.array(r_distribution1).mean(axis = 0)
  r_distribution2 = np.array(r_distribution2).mean(axis = 0)
  r_distribution3 = np.array(r_distribution3).mean(axis = 0)

  r_cumulative1 = np.array(r_cumulative1).mean(axis = 0)
  r_cumulative2 = np.array(r_cumulative2).mean(axis = 0)
  r_cumulative3 = np.array(r_cumulative3).mean(axis = 0)


  p_distribution1 = np.array(p_distribution1).mean(axis = 0)
  p_distribution2 = np.array(p_distribution2).mean(axis = 0)
  p_distribution3 = np.array(p_distribution3).mean(axis = 0)

  p_cumulative1 = np.array(p_cumulative1).mean(axis = 0)
  p_cumulative2 = np.array(p_cumulative2).mean(axis = 0)
  p_cumulative3 = np.array(p_cumulative3).mean(axis = 0)


  r_mean1= np.array(r_mean1).mean()
  r_mean2= np.array(r_mean2).mean()
  r_mean3= np.array(r_mean3).mean()

  p_mean1= np.array(p_mean1).mean()
  p_mean2= np.array(p_mean2).mean()
  p_mean3= np.array(p_mean3).mean()
  
  output.put((r_distribution1, r_distribution2, r_distribution3, r_cumulative1, r_cumulative2, r_cumulative3, p_distribution1, p_distribution2, p_distribution3, p_cumulative1, p_cumulative2, p_cumulative3,
  r_mean1, r_mean2, r_mean3, p_mean1, p_mean2, p_mean3))