time_num = settings.TIME_SLICE
train_percent = 0.95
alpha = 0.95
x_values = []
y_values1 = []
y_values2 = []
y_values3 = []
# data, axis_users, axis_pois, check_data = init_data(region, filter_count, train_percent)
data, axis_users, axis_pois, check_data = init_data_by_user(tuple([0, 3, 4, 5, 30]), filter_count, train_percent)
user_num = len(axis_users)
poi_num = len(axis_pois)
init_tensor = [[[1/(poi_num * time_num * user_num) for i in range(poi_num)] for j in range(time_num)] for k in range(user_num)]
transition_tensor = inreducible_tensor(mtt(data, user_num, poi_num), user_num, time_num, poi_num, alpha)
res1, iterator_values1 = tensor_three_mode_product(transition_tensor, init_tensor)
order = 1
while order <= settings.ITERATOR_NUMBER:
x_values.append(order)
y_values1.append(iterator_values1[order])
order += 1
label = ""+str(((settings.TIME_SLICE*user_num*poi_num)**2))
pylab.plot(x_values, y_values1, 'rs', linewidth=1, linestyle="-", label=label)
data, axis_users, axis_pois, check_data = init_data_by_user(tuple([0, 3, 4]), filter_count, train_percent)
user_num = len(axis_users)
poi_num = len(axis_pois)
if __name__ == '__main__':
# beijing = (39.433333, 41.05, 115.416667, 117.5)
# haidian = (39.883333, 40.15, 116.05, 116.383333)
region = (39.883333, 40.05, 116.05, 116.383333)
filter_count = 600
alpha = 0.8
alpha_shift = 0.1
data, axis_users, axis_pois, check_data = init_data(region, filter_count)
user_num = len(axis_users)
time_num = settings.TIME_SLICE
poi_num = len(axis_pois)
transition_tensor = mtt(data, user_num, poi_num)
transition_tensor2 = inreducible_tensor(transition_tensor, user_num, time_num, poi_num, alpha)
transition_tensor3 = mtt(data, user_num, poi_num, zero_adjustment=False)
# equal_all_sum_one: equal
# init_tensor1 = [[[1/(poi_num * time_num * user_num) for i in range(poi_num)] for j in range(time_num)] for k in range(user_num)]
# # random_all_sum_one: no zero element
# temp_tensor = [[[random.choice([1, 2, 3, 100]) for i in range(poi_num)] for j in range(time_num)] for k in range(user_num)]
# init_tensor2 = three_tensor_hadarmard(1/three_order_tensor_first_norm(temp_tensor), temp_tensor)
#
# init_tensor3 = [[[0 for i in range(poi_num)] for j in range(time_num)] for k in range(user_num)]
# init_tensor3[0][0][0] = 1
# # random_all_sum_one: exist zero element
# temp_tensor = [[[random.choice([0, 1, 2, 3, 4]) for i in range(poi_num)] for j in range(time_num)] for k in range(user_num)]
# init_tensor3 = three_tensor_hadarmard(1/three_order_tensor_first_norm(temp_tensor), temp_tensor)
# # user_slice_sum_one: equal
filter_count = 600
time_num = settings.TIME_SLICE
train_percent = 0.1
x_values = []
y_values1 = []
y_values2 = []
order = 1
while train_percent < 1:
data, axis_users, axis_pois, check_data = init_data(
region, filter_count, train_percent)
user_num = len(axis_users)
poi_num = len(axis_pois)
transition_tensor = mtt(data, user_num, poi_num)
# equal_all_sum_one: equal
init_tensor1 = [[[
1 / (poi_num * time_num * user_num) for i in range(poi_num)
] for j in range(time_num)] for k in range(user_num)]
# user_slice_sum_one: equal
init_tensor2 = [[[1 / (poi_num * time_num) for i in range(poi_num)]
for j in range(time_num)] for k in range(user_num)]
res1, iterator_values1 = tensor_three_mode_product(
transition_tensor, init_tensor1)
res2, iterator_values2 = tensor_three_mode_product(
transition_tensor, init_tensor2)
# print res
values = []
x_values = []
y_values1 = []
y_values2 = []
y_values3 = []
y_values4 = []
order = 1
while alpha >= 0.2:
data, axis_users, axis_pois, check_data = init_data(
region, filter_count, train_percent)
user_num = len(axis_users)
poi_num = len(axis_pois)
transition_tensor = inreducible_tensor(mtt(data, user_num,
poi_num), user_num,
time_num, poi_num, alpha)
# equal_all_sum_one: equal
init_tensor = [[[
1 / (poi_num * time_num * user_num) for i in range(poi_num)
] for j in range(time_num)] for k in range(user_num)]
res, iterator_values = tensor_three_mode_product(
transition_tensor, init_tensor)
values.append(iterator_values)
alpha -= 0.2
print values
# print values.keys()
value1 = values[0]
region = (39.883333, 40.05, 116.05, 116.383333)
filter_count = 600
time_num = settings.TIME_SLICE
train_percent = 0.1
x_values = []
y_values1 = []
y_values2 = []
order = 1
while train_percent < 1:
data, axis_users, axis_pois, check_data = init_data(region, filter_count, train_percent)
user_num = len(axis_users)
poi_num = len(axis_pois)
transition_tensor = mtt(data, user_num, poi_num)
# equal_all_sum_one: equal
init_tensor1 = [[[1/(poi_num * time_num * user_num) for i in range(poi_num)] for j in range(time_num)] for k in range(user_num)]
# user_slice_sum_one: equal
init_tensor2 = [[[1/(poi_num * time_num) for i in range(poi_num)] for j in range(time_num)] for k in range(user_num)]
res1, iterator_values1 = tensor_three_mode_product(transition_tensor, init_tensor1)
res2, iterator_values2 = tensor_three_mode_product(transition_tensor, init_tensor2)
# print res
check_tensor = get_check_tensor(check_data, user_num, time_num, poi_num)
residual1 = delta_tensor_norm(res1, check_tensor)
residual2 = delta_tensor_norm(res2, check_tensor)