barrel = []
for index_ in range(len(u2d_adj_close_index) - u2d_split_length):
feature_point_index_0, feature_point_index_1 = u2d_adj_close_index[
index_], u2d_adj_close_index[index_ + u2d_split_length]
feature_point_values = train_data[
feature_point_index_0:feature_point_index_1]
barrel.append(feature_point_values)
barrel_length = len(barrel)
nn_distance = np.zeros([barrel_length, barrel_length], dtype=np.float)
distance_list = []
for i_ in range(barrel_length - 1):
for j_ in range(i_ + 1, barrel_length):
dist, cost, acc, path = dtw(barrel[i_],
barrel[j_],
dist=lambda x, y: abs(x - y))
nn_distance[i_, j_] = dist
nn_distance[j_, i_] = dist
distance_list.append(dist)
distance_array = np.array(distance_list)
cutoff_position = int(len(distance_array) * cutoff_present)
sorted_arr = np.sort(distance_array)
cutoff_distance = sorted_arr[cutoff_position]
rho = np.zeros(barrel_length, dtype=np.float)
for i in range(barrel_length - 1):
for j in range(i, barrel_length):
dist = np.exp(-(nn_distance[i][j] / cutoff_distance)**2)
rho[i] += dist
prices = pd.read_csv("../data/^GSPC.csv")
adj_close = prices.iloc[:, 3]
index = prices.index
start = 0
offset = 100
a1 = adj_close[start: start+offset].values
start = 300
a2 = adj_close[start: start+offset].values
dist = lambda x, y: np.linalg.norm(np.array(x)-np.array(y), ord=2)
# res, path, acc = DTW(a1, a2, fun=dist)
import math
res, cost, acc, path = dtw(np.array(a1).reshape(-1, 1), np.array(a2).reshape(-1, 1), dist=lambda x, y: abs(x-y))
print("res1", res)
dist2, cost2, acc2, path2 = thedtw.dtw(np.array(a1).reshape(-1, 1), np.array(a2).reshape(-1, 1), dist=lambda x, y: np.linalg.norm(x - y, ord=1))
print("res2", dist2)
print("acc1")
print(acc)
print("acc2")
print(acc2)
print("path1")
print(path)
print("path2")
print(path2)