def GA_const_JandH(w, C_JandH, alpha_JandH, ngen_count, feasible, adjuster):
xtvar_check = Ityt.Uwg(ITMC.MCbase(Premium_list, w_initial, cycle), ITMC.MCbase(Loss_list, w_initial, cycle),
ITMC.MCbase(Expense_list, w_initial, cycle)).xtvar(alpha) - Ityt.Uwg(
ITMC.MCbase(Premium_list, w, cycle), ITMC.MCbase(Loss_list, w, cycle),
ITMC.MCbase(Expense_list, w, cycle)).xtvar(alpha)
if abs(xtvar_check) < feasible:
feasible_ret = 0
else:
feasible_ret = (C_JandH * ngen_count) ** alpha_JandH * abs(xtvar_check)
return feasible_ret * adjuster
model_TVaR.setObjective(
quicksum(
Aggregate(x[i], Premium_list_gurobi[i], realization_num) -
Aggregate(x[i], Loss_list_gurobi[i], realization_num) -
Aggregate(x[i], Expense_list_gurobi[i], realization_num)
for i in range(I)) / realization_num, GRB.MAXIMIZE)
model_TVaR.update()
model_TVaR.__data = x
model_TVaR.__data = VaR
return model_TVaR
default_TVaR = Ityt.Uwg(ITMC.MCbase(Premium_list, w0, cycle),
ITMC.MCbase(Loss_list, w0, cycle),
ITMC.MCbase(Expense_list, w0, cycle)).xtvar(alpha)
model_TVaR = markowitz_TVaR(I, cycle, default_TVaR, alpha, Premium_list_gurobi,
Loss_list_gurobi, Expense_list_gurobi)
model_TVaR.optimize()
status = model_TVaR.Status
if status == GRB.Status.UNBOUNDED or status == GRB.Status.INF_OR_UNBD:
model_TVaR.setObjective(0, GRB.MAXIMIZE)
model_TVaR.optimize()
status = model_TVaR.Status
if status == GRB.Status.OPTIMAL:
print 'UNbounded'
elif status == GRB.Status.INFEASIBLE:
def obj_func_ga_initial(w1):
return Ityt.Uwg(ITMC.MCbase(Premium_list, w1, cycle), ITMC.MCbase(Loss_list, w1, cycle),
ITMC.MCbase(Expense_list, w1, cycle)).rorac(alpha),
def obj_func_ga(w1, ngen_count):
return Ityt.Uwg(ITMC.MCbase(Premium_list, w1, cycle), ITMC.MCbase(Loss_list, w1, cycle),
ITMC.MCbase(Expense_list, w1, cycle)).rorac(alpha) - GA_const_JandH(w1, C_JandH,
alpha_JandH,
ngen_count, feasible,
adjuster),
def obj_func_PSO(w1):
w1 = list(np.array(w1) * np.array(alpso_weight) + 1.)
f = Ityt.Uwg(ITMC.MCbase(Premium_list, w1, cycle), ITMC.MCbase(Loss_list, w1, cycle),
ITMC.MCbase(Expense_list, w1, cycle)).rorac(alpha) * -1.
g = [0.] * 2
g[0] = Ityt.Uwg(ITMC.MCbase(Premium_list, w1, cycle), ITMC.MCbase(Loss_list, w1, cycle),
ITMC.MCbase(Expense_list, w1, cycle)).xtvar(alpha) - Ityt.Uwg(
ITMC.MCbase(Premium_list, w_initial, cycle), ITMC.MCbase(Loss_list, w_initial, cycle),
ITMC.MCbase(Expense_list, w_initial, cycle)).xtvar(alpha)
g[1] = 0.9 * Ityt.Uwg(ITMC.MCbase(Premium_list, w_initial, cycle), ITMC.MCbase(Loss_list, w_initial, cycle),
ITMC.MCbase(Expense_list, w_initial, cycle)).xtvar(alpha) - Ityt.Uwg(
ITMC.MCbase(Premium_list, w1, cycle), ITMC.MCbase(Loss_list, w1, cycle),
ITMC.MCbase(Expense_list, w1, cycle)).xtvar(alpha)
fail = 0
print 'obj_value : ', f, ' variable : ', w1
print 'penalty(negative value is valid) : ', g[0], g[1]
return f, g, fail
def obj_func(w1):
return (Ityt.Uwg(ITMC.MCbase(Premium_list, w1, cycle), ITMC.MCbase(Loss_list, w1, cycle),
ITMC.MCbase(Expense_list, w1, cycle)).uwg_mean() - (Risk_charge * np.array(w1)).sum()) * -1. / total_capital
'skyblue', 'crimson']
#######################################################################################################################
#
# todo parameter setting
work_dir_name = '1512macro/'
accumulated_dir = './accumulated_data/' + work_dir_name # all data is class type
accumulated_dir_risk_metric = './risk_metrics/' + work_dir_name
global_opt_result = './global_opt_result/' + work_dir_name
local_opt_result = './local_opt_result/' + work_dir_name
cycle = 100000
alpha = 0.995 # 200yr --> 0.995 100yr --> 0.99
beta = 0.998
MC_cycle = 2000
num_clusters = 5
global_const = [lambda w: ITMC.MCbase(USEQ_list, w, cycle).tvar(alpha) / SGO_USEQ_limit < 1.1,
lambda w: ITMC.MCbase(USWS_list, w, cycle).tvar(alpha) / SGO_USWS_limit < 1.1,
lambda w: ITMC.MCbase(EUWS_list, w, cycle).tvar(alpha) / SGO_EUWS_limit < 1.1,
lambda w: ITMC.MCbase(JPWS_list, w, cycle).tvar(alpha) / SGO_JPWS_limit < 1.1,
lambda w: ITMC.MCbase(JPEQ_list, w, cycle).tvar(alpha) / SGO_JPEQ_limit < 1.1,
lambda w: ITMC.MCbase(RoW_list, w, cycle).tvar(alpha) / SGO_RoW_limit < 1.1
]
weight_list = list(pd.read_table('weight.txt').columns) # setting of Optimized Variable's name
w_initial = [1.] * len(weight_list)
w_lower = list(pd.read_table('weight.txt').ix[0, :])
w_upper = list(pd.read_table('weight.txt').ix[1, :])
bnds = np.array([w_lower, w_upper])
Premium_list = []
Loss_list = []
Expense_list = []