Question

我试图获得两个时间序列与DTW之间的相关性。但我发现时间序列的幅度会影响结果。这是我的代码：

import numpy as np
import rpy2.robjects.numpy2ri
rpy2.robjects.numpy2ri.activate()
from rpy2.robjects.packages import importr

# Set up our R namespaces
R = rpy2.robjects.r
DTW = importr('dtw')

L1 = [np.sin(i*.01) for i in range(350)];
L2 = [np.sin(i*.01) + 1 for i in range(350)]
L3 = 100*[np.sin(i*.01) + 1 for i in range(350)]


alignment = R.dtw(L1, L2, keep=True)
correlation1 = alignment.rx('normalizedDistance')[0][0]

alignment = R.dtw(L1, L3, keep=True)
correlation2 = alignment.rx('normalizedDistance')[0][0]

correlation1的值为0.4427365468841718，correlation2的值为0.5861839240861364。

我想知道是否有一种简单的方法来规范化结果。提前致谢

Answer 1

一种选择是自己规范化时间序列。这使得流程移动和缩放不变。

import numpy as np
import matplotlib.pyplot as plt

def normalise(series):
    max_value = max(series)
    min_value = min(series)
    return (series - min_value) / (max_value - min_value)

if __name__ == "__main__":
    L1 = [np.sin(i*.01) for i in range(350)]
    L2 = [np.sin(i*.01) + 1 for i in range(350)]
    L3 = 100*[np.sin(i*.01) + 1 for i in range(350)]

    norm_L1 = normalise(L1)
    norm_L2 = normalise(L2)
    norm_L3 = normalise(L3)

    # Correlate the normalised signals
    alignment = R.dtw(norm_L1, norm_L2, keep=True)
    correlation1 = alignment.rx('normalizedDistance')[0][0]

    alignment = R.dtw(norm_L1, norm_L3, keep=True)
    correlation2 = alignment.rx('normalizedDistance')[0][0]

现在norm_L1/L2/L3在0和1之间都有所不同。 norm_L2和norm_L3实际上是相同的，因此它们与norm_L1的相关性保证相同。

动态时间包（DTW）规范化，python实现

1 个答案: