def run_qp(mean_vec, cov_mat, alpha, k, separate_profit_vec):
print('testing the QP relaxation algorithm...')
start_time = time.time()
bundle_set, profit_bundle = kbundle.Kbundle_QP_relaxation(
mean_vec, cov_mat, alpha, k, separate_profit_vec)
running_time = time.time() - start_time
return profit_bundle, bundle_set, running_time
def run_qp(mean_vec, cov_mat, alpha, k, separate_profit_vec):
print('Testando algoritmo de relaxation QP')
start_time = time.time()
bundle_set, profit_bundle = kbundle.Kbundle_QP_relaxation(
mean_vec, cov_mat, alpha, k, separate_profit_vec)
running_time = time.time() - start_time
return profit_bundle, bundle_set, running_time
print("Parte 4")
start = time.time()
std_array = np.asarray(std_vec).reshape((len(std_vec), 1))
std_scale_mat = np.dot(std_array, np.transpose(std_array))
cov_mat = np.multiply(predicted_cov_mat, std_scale_mat).astype(np.double)
alpha = search(sales_vec)
# precompute the separate sale profit
separate_profit_vec = profit.separate_sale_profit(mean_vec, cov_mat, alpha)
profit_vec = []
bundle_size_vec = [2]
# bundle_size_vec_norm = np.where(np.asarray(bundle_size_vec) >= N_mandatory_products)
for bundle_size in bundle_size_vec:
bundle_set, profit_bundle = kbundle.Kbundle_QP_relaxation(mean_vec, cov_mat, alpha, bundle_size,
separate_profit_vec, mandatory_products_indexes)
profit_vec.append(profit_bundle)
with open('data/result_kbundle_' + str(bundle_size) + '.json', 'w') as output_file:
json.dump({"profit": profit_bundle, "set": list(bundle_set)}, output_file, indent=4)
with open('data/result_kbundle_profits.json', 'w') as output_file:
json.dump({"bundle_size_vec": bundle_size_vec, "profit_vec": profit_vec}, output_file, indent=4)
tempos_execucao[3] = time.time() - start
max_products_sold = int(customer_product_reduced['sold_units'].max())
products_sold_axis = np.linspace(1, max_products_sold, max_products_sold)
count_sales = customer_product_reduced.groupby('sold_units').count()
###############################################################################
print("Parte 4")
start = time.time()
std_array = np.asarray(std_vec).reshape((len(std_vec), 1))
std_scale_mat = np.dot(std_array, np.transpose(std_array))
cov_mat = np.multiply(predicted_cov_mat, std_scale_mat).astype(np.double)
alpha = search(sales_vec)
# precompute the separate sale profit
separate_profit_vec = profit.separate_sale_profit(mean_vec, cov_mat, alpha)
profit_vec = []
bundle_size_vec = [2]
for bundle_size in bundle_size_vec:
bundle_set, profit_bundle = kbundle.Kbundle_QP_relaxation(
mean_vec, cov_mat, alpha, bundle_size, separate_profit_vec)
profit_vec.append(profit_bundle)
with open('data/result_kbundle_' + str(bundle_size) + '.json',
'w') as output_file:
json.dump({
"profit": profit_bundle,
"set": list(bundle_set)
},
output_file,
indent=4)
with open('data/result_kbundle_profits.json', 'w') as output_file:
json.dump(
{
"bundle_size_vec": bundle_size_vec,
for N in N_vec:
print ("N =", N)
for k in k_vec:
print (">> k =", k)
profit_algo_vec = []
arg_vec = []
running_time_algo_vec = []
for s in range(sample_size):
rand_indices = list(np.random.choice(N_product, N))
mean_vec_sub = [mean_vec[i] for i in rand_indices]
cov_mat_sub = cov_mat[:, rand_indices][rand_indices]
separate_profit_vec_sub = [separate_profit_vec[i] for i in rand_indices]
start_time = time.time()
_, profit_algo = kbundle.Kbundle_QP_relaxation(mean_vec_sub, cov_mat_sub, alpha, k, separate_profit_vec_sub)
running_time_algo = time.time() - start_time
running_time_algo_vec.append(running_time_algo)
arg_vec.append((mean_vec_sub, cov_mat_sub, k, separate_profit_vec_sub))
profit_algo_vec.append( float(profit_algo) )
print ('N = ', N, 'k = ', k)
start_time = time.time()
profit_opt_vec = pool.starmap(opt_enumerate, arg_vec)
profit_opt_vec = list(profit_opt_vec)
running_time_enumerate = time.time() - start_time
ratio_vec = list(map(lambda x,y:x/y, profit_algo_vec, profit_opt_vec))