def speed_est(road, v_diff_temp, theta, lamda, Tcon, v_diff_):
if road in SEED:
return -1
#global Un,Ue
Un, Ue = traffic_trend.create_graph(road)
UnS = [val for val in Un if val in SEED]
A = list(set(Un) & set(all_road_info[road][-1]))
AS = [val for val in A if val in SEED]
v_equal = all_road_info[road][4]
delta_v = traffic_trend.trend_infer(road, Un, Ue)
A_S = [val for val in A if val not in SEED]
#W = get_weight({},road,Un,Ue)
#print W
if len(AS) == 0:
#v_diff_temp[road][date] = 0
return v_equal
elif len(AS) >= len(A_S):
Un = UnS
A = AS
#W = weight_learning(road,Un,Ue,W,theta,lamda,Tcon,v_diff_temp)
else:
for road1 in A_S:
Un1, _ = traffic_trend.create_graph(road1) #???bug
A1 = list(set(Un1) & set(all_road_info[road][-1]))
#print Un1
A1S = list(set(A1) & set(SEED))
A1_S = list(set(A1) - set(SEED))
if len(A1S) == 0 or len(A1S) < len(A1_S): #1表示road1本身
Un = list(set(Un) - set([road1]))
A = list(set(A) - set([road1]))
W = get_weight({}, road, Un, Ue)
#print W
if len(UnS) == 0:
return v_equal
W = weight_learning(road, Un, Ue, W, theta, lamda, Tcon, v_diff_)
v_diff = speed_diff(W, road, Un, Ue, v_diff_temp)
v_diff_temp[road] = v_diff
#v_diff = speed_diff(W,road,Un,Ue,v_diff_temp)
#v_diff_temp[road] = v_diff
#print v_diff
if delta_v > 0:
v_est = v_equal + v_diff
else:
v_est = v_equal - v_diff
return v_est
def speed_diff(W, road, Un, Ue, v_diff):
#1-hop相关邻居
A = list(set(Un) & set(all_road_info[road][-1]))
AS = [val for val in A if val in SEED]
#W = get_weight(road)
#v_diff = V_diff(road,date)
diff_est = {}
res = 0
A_S = [val for val in A if val not in SEED]
#W = get_weight({},road,Un,Ue)
if len(AS) == 0:
v_diff[road] = 0
elif len(AS) >= len(set(A_S)):
A = AS
#W = weight_learning(road,Un,Ue,W,theta,lamda,Tcon,v_diff_temp)
else:
for road1 in A_S:
Un1, _ = traffic_trend.create_graph(road1) #???bug
#print Un1
A1 = list(set(Un1) & set(all_road_info[road1][-1]))
A1S = list(set(A1) & set(SEED))
A1_S = list(set(A1) - set(SEED))
if len(A1S) == 0 or len(A1S) < len(A1_S): #1表示road1本身
Un = list(set(Un) - set([road1]))
A = list(set(A) - set([road1]))
#W = weight_learning(road1,Un1,Ue1,W,theta,lamda,Tcon,v_diff_temp)
#W = weight_learning(road,Un,Ue,W,theta,lamda,Tcon,v_diff_temp)
#UnS = [val for val in Un if val in SEED]
#A = list(set(Un)&set(all_road_info[road][-1]))
A_S = [val for val in A if val not in SEED]
UnS = [val for val in Un if val in SEED]
for road_j in A:
if road_j in SEED:
#print road_j
diff_est[road_j] = v_diff[road_j]
else:
temp = 0
for road_l in UnS:
if road_l + '-' + road_j in W.keys():
temp += W[road_l + '-' + road_j] * v_diff[road_l]
diff_est[road_j] = temp
#print v_diff
#print diff_est
#print W
#A会发生改变?
for road_j in A:
res += W[road_j] * diff_est[road_j]
return res
def V_diff(road,date):
Un,Ue = traffic_trend.create_graph(road)
UnS = [val for val in Un if val in SEED]
v_diff = {}
for road_l in UnS:
if date in c_weekday_d:
v_equal = all_road_info[road_l][4]
v = all_road_info[road_l][0][date]
v_diff[road_l] = abs(v-v_equal)
if date in c_weekend_d:
v_equal = all_road_info[road_l][5]
v = all_road_info[road_l][2][date]
v_diff[road_l] = abs(v-v_equal)
return v_diff
def get_weight(W,road,Un,Ue):
#Un,Ue = traffic_trend.create_graph(road)
#A = [val for val in Un if val != road]
A = list(set(Un)&set(all_road_info[road][-1]))
UnS = [val for val in Un if val in SEED]
AS = [val for val in A if val in SEED]
A_S = [val for val in A if val not in SEED]
#print A
#print UnS
if len(AS) == 0:
print '111'
return W
elif len(AS) >= len(set(A_S)):
print '222'
Un = UnS
A = AS
#W = weight_learning(road,Un,Ue,W,theta,lamda,Tcon,v_diff_temp)
else:
print '333'
for road1 in A_S:
Un1,Ue1 = traffic_trend.create_graph(road1) #???bug
#print Un1
A1 = list(set(Un1)&set(all_road_info[road1][-1]))
A1S = list(set(A1)&set(SEED))
A1_S = list(set(A1)-set(SEED))
if len(A1S) == 0 or len(A1S) < len(A1_S): #1表示road1本身
Un = list(set(Un)-set([road1]))
A = list(set(A)-set([road1]))
A_S = [val for val in A if val not in SEED]
UnS = [val for val in Un if val in SEED]
for road1 in A:
W[road1] = random()
for road_j in A_S:
for road_l in UnS:
if [road_l,road_j] in Ue or [road_j,road_l] in Ue:
#格式road1-road2
W[road_l+'-'+road_j] = random()
print 'w-----'
print A
print UnS
print W
print road
return W
def speed_est(road,v_diff_temp,theta,lamda,Tcon,v_diff_):
if road in SEED:
print "choose another non-seed road"
return -1
#global Un,Ue
Un,Ue = traffic_trend.create_graph(road)
print Un
UnS = [val for val in Un if val in SEED]
A = list(set(Un)&set(all_road_info[road][-1]))
AS = [val for val in A if val in SEED]
print A
v_equal = all_road_info[road][4]
delta_v = traffic_trend.trend_infer(road,Un,Ue)
A_S = [val for val in A if val not in SEED]
#W = get_weight({},road,Un,Ue)
#print W
if len(AS) == 0:
print '1'
#v_diff_temp[road][date] = 0
return v_equal
elif len(AS) >= len(A_S):
print '2'
Un = UnS
A = AS
#W = weight_learning(road,Un,Ue,W,theta,lamda,Tcon,v_diff_temp)
else:
print '3'
for road1 in A_S:
Un1,Ue1 = traffic_trend.create_graph(road1) #???bug
A1 = list(set(Un1)&set(all_road_info[road][-1]))
#print Un1
A1S = list(set(A1)&set(SEED))
A1_S = list(set(A1)-set(SEED))
if len(A1S) == 0 or len(A1S) < len(A1_S): #1表示road1本身
print road1
Un = list(set(Un)-set([road1]))
A = list(set(A)-set([road1]))
print A
print A
W = get_weight({},road,Un,Ue)
#print W
if len(UnS) == 0:
return v_equal
print "2----"
print A
print Un
print W
W = weight_learning(road,Un,Ue,W,theta,lamda,Tcon,v_diff_)
print 'W------'
print W
v_diff = speed_diff(W,road,Un,Ue,v_diff_temp)
v_diff_temp[road] = v_diff
#v_diff = speed_diff(W,road,Un,Ue,v_diff_temp)
#v_diff_temp[road] = v_diff
#print v_diff
if delta_v > 0:
v_est = v_equal + v_diff
else:
v_est = v_equal - v_diff
return v_est
def weight_learning(road,Un,Ue,W,theta,lamda,Tcon,v_diff_):
#W0 = get_weight(W,road,Un,Ue)
V_diff_road = {}
#V_diff_UnS = {}
V_diff_est = {}
#v_diff_final = {}
A = [val for val in Un if val != road]
UnS = [val for val in Un if val in SEED]
for date in c_weekday_d:
v_equal = all_road_info[road][4]
v = all_road_info[road][0][date]
V_diff_road[date] = abs(v-v_equal)
#V_diff_UnS[road][date] = V_diff(road_uns,date)
#print V_diff_road
#print W
v_diff_sum = 0
for date in c_weekday_d:
#v_diff = V_diff(road,date)
V_diff_est[date] = speed_diff(W,road,Un,Ue,v_diff_)
v_diff_sum += V_diff_est[date]
v_diff_new = v_diff_sum/len(c_weekday_d)
i = 0
while 1:
v_diff_org = v_diff_new
#print v_diff_sum
#print v_diff_org
#theta = 0.09
#lamda = 1
#Tcon = 0.001
for value in W:
if '-' not in value:
#v_diff_j = {}
#if 1:
sum_j = 0
for date in c_weekday_d:
#v_equal_j = all_road_info[value][4]
#v_j = all_road_info[value][0][date]
if value in v_diff_.keys() and v_diff_[value] != 0 :
v_diff_j = v_diff_[value]
else:
Un_value,Ue_value = traffic_trend.create_graph(value)
'''
#W0 = get_weight(W0,value,Un_value,Ue_value)
W_value = weight_learning(value,Un_value,Ue_value,W0,theta,lamda,Tcon)
v_diff_value = V_diff(value,date)
v_diff_j = speed_diff(W_value,value,Un_value,Ue_value,v_diff_value)
v_diff_final[value][date] = v_diff_j'''
W_value = get_weight({},value,Un_value,Ue_value)
#嵌套循环?
#W_value = weight_learning(value,Un_value,Ue_value,W_value,theta,lamda,Tcon,v_diff_)
v_diff_j = speed_diff(W_value,value,Un_value,Ue_value,v_diff_)#abs(speed_est(W,v_diff_,theta,lamda,Tcon)-all_road_info[value][0][date])
print "loop"
sum_j += (V_diff_est[date]-V_diff_road[date])*v_diff_j
W[value] -= theta*(sum_j/len(c_weekday_d)+lamda*W[value])
if W[value] < 0:
W[value] = 0
if W[value] > 1:
W[value] = 1
'''for road_j in A:
if [value,road_j] not in W.keys() or [road_j,value] not in W.keys():
continue'''
if '-' in value:
#print value
#v_diff_j = {}
sum_lj = 0
road_l = value.split('-')[0]
road_j = value.split('-')[1]
for date in c_weekday_d:
v_equal_l = all_road_info[road_l][4]
v_l = all_road_info[road_l][0][date]
#Un_value,Ue_value = traffic_trend.create_graph(value)
v_diff_l = abs(v_l-v_equal_l)
sum_lj += (V_diff_est[date]-V_diff_road[date])*v_diff_l*W[road_j]
W[value] -= theta*(sum_lj/len(c_weekday_d)+lamda*W[value])
if W[value] < 0:
W[value] = 0
if W[value] > 1:
W[value] = 1
v_diff_sum = 0
for date in c_weekday_d:
V_diff_est[date] = speed_diff(W,road,Un,Ue,v_diff_)
v_diff_sum += V_diff_est[date]
v_diff_new = v_diff_sum/len(c_weekday_d)
#print v_diff_new
i += 1
if abs(v_diff_new-v_diff_org) < Tcon or i > 200:
#print "-----"
break
return W
if int(record[9]) - int(period) == -1:
temp_v = float(record[6])
elif record[9] == period:
v = float(record[6])
temp_v = 0
elif int(record[9]) - int(period) == 1 and temp_v != 0 and float(record[6]) - temp_v < 4:
v = (temp_v + float(record[6]))/2
v_temp1[road1] += v
if v != 0:
date_num += 1
print date_num
v_equal1[road1] = v_temp1[road1]/date_num
#记录非种子路段的Un中的种子个数
for road in non_seed_road:
Un1,Ue1 = traffic_trend.create_graph(road)
A1 = list(set(Un1)&set(all_road_info[road][-1]))
A1S = [val for val in A1 if val in SEED]
seed_num[road] = len(A1S)
'''
value = list(seed_num.values()).sort(reverse = True)
for i in value:
index1 = seed_num.values().index(i)
NSR_order.append(seed_num.keys()[index1])
'''
n= 0
#Un无种子
for road in non_seed_road:
Un1,Ue1 = traffic_trend.create_graph(road)
UnS1 = [val for val in Un1 if val in SEED]
A1 = list(set(Un1)&set(all_road_info[road][-1]))