minirocket 数据维度

125404629

光谱数据维度有855个样本，890个变量，输入minirocket出现错误。
X.shape=(855,890),Y.shape=(855,1)

发生异常: ValueError
n_timepoints must be >= 9, but found 1; zero pad shorter series so that n_timepoints == 9
  File "D:\0000可见光2\程序\MiniRocket\Test1.py", line 20, in <module>
    model.fit(X_train, y_train)
ValueError: n_timepoints must be >= 9, but found 1; zero pad shorter series so that n_timepoints == 9

1. from tsai.all import MiniRocket
   
2. import numpy as np
   
3. import pandas as pd
   
4. from sklearn.model_selection import train_test_split
   
5. from tsai.basics import *
   
6. from tsai.models.MINIROCKET import MiniRocketClassifier
   
7. 
   
8. df1 = pd.read_csv("train.csv")
   
9. df1 = np.array(df1)
   
10. X = df1[:, 1:891]
    
11. X = np.expand_dims(df1[:, 1:891].astype(float), axis=2)
    
12. Y = df1[:, 0]
    
13. print(X.shape, Y.shape)
    
14. df2 = pd.read_csv("test.csv")
    
15. df2 = np.array(df1)
    
16. X_test = df2[:, 1:891]
    
17. X_train, X_valid, y_train, y_valid = train_test_split(X, Y, test_size=0.8, random_state=42)
    
18. print(X_train.shape, y_train.shape, X_valid.shape, y_valid.shape)
    
19. model = MiniRocketClassifier()
    
20. model.fit(X_train, y_train)
    
21. 
    
22. cls = MiniRocket()
    
23. y=model.score(X_valid, y_valid)
    
24. print(y)

复制代码

isdkz

这个错误是因为 MiniRocket 要求输入的时间序列长度至少为 9，而您的数据长度只有 1，因此需要使用零填充将其长度增加到 9。您可以使用 tsai 库中的 pad_min_len 函数来实现这一点。以下是修改后的代码：

1. from tsai.all import MiniRocket, pad_min_len
   
2. import numpy as np
   
3. import pandas as pd
   
4. from sklearn.model_selection import train_test_split
   
5. from tsai.basics import *
   
6. from tsai.models.MINIROCKET import MiniRocketClassifier
   
7. 
   
8. df1 = pd.read_csv("train.csv")
   
9. df1 = np.array(df1)
   
10. X = df1[:, 1:891]
    
11. X = np.expand_dims(df1[:, 1:891].astype(float), axis=2)
    
12. Y = df1[:, 0]
    
13. print(X.shape, Y.shape)
    
14. df2 = pd.read_csv("test.csv")
    
15. df2 = np.array(df1)
    
16. X_test = df2[:, 1:891]
    
17. 
    
18. # 使用 pad_min_len 函数将 X 的长度增加到 9
    
19. X = pad_min_len(X, 9)
    
20. 
    
21. X_train, X_valid, y_train, y_valid = train_test_split(X, Y, test_size=0.8, random_state=42)
    
22. print(X_train.shape, y_train.shape, X_valid.shape, y_valid.shape)
    
23. model = MiniRocketClassifier()
    
24. model.fit(X_train, y_train)
    
25. 
    
26. cls = MiniRocket()
    
27. y = model.score(X_valid, y_valid)
    
28. print(y)
    

复制代码

希望能够帮到您！

陶远航

您的问题似乎是关于使用MiniRocket模型处理时间序列数据时遇到的问题。MiniRocket是一个高效的时间序列分类模型，它需要输入的数据是一个二维数组，其中每一行是一个时间序列样本，每一列是一个时间点。

在您的代码中，您使用了np.expand_dims(df1[:, 1:891].astype(float), axis=2)来增加一个维度，这可能是问题的原因。这一步将数据从二维转换为三维，可能导致MiniRocket无法正确处理。

另外，错误信息n_timepoints must be >= 9, but found 1; zero pad shorter series so that n_timepoints == 9表明，MiniRocket期望每个时间序列样本至少有9个时间点，但在您的数据中只找到了1个。这可能是因为增加了一个不必要的维度导致的。

您可以尝试以下修改：

1. from tsai.all import MiniRocket
   
2. import numpy as np
   
3. import pandas as pd
   
4. from sklearn.model_selection import train_test_split
   
5. from tsai.basics import *
   
6. from tsai.models.MINIROCKET import MiniRocketClassifier
   
7. 
   
8. df1 = pd.read_csv("train.csv")
   
9. df1 = np.array(df1)
   
10. X = df1[:, 1:891].astype(float)  # 不增加额外的维度
    
11. Y = df1[:, 0]
    
12. print(X.shape, Y.shape)
    
13. 
    
14. df2 = pd.read_csv("test.csv")
    
15. df2 = np.array(df1)
    
16. X_test = df2[:, 1:891]
    
17. 
    
18. X_train, X_valid, y_train, y_valid = train_test_split(X, Y, test_size=0.8, random_state=42)
    
19. print(X_train.shape, y_train.shape, X_valid.shape, y_valid.shape)
    
20. 
    
21. model = MiniRocketClassifier()
    
22. model.fit(X_train, y_train)
    
23. 
    
24. cls = MiniRocket()
    
25. y=model.score(X_valid, y_valid)
    
26. print(y)
    

复制代码

这样，您的数据应该能够正确地被MiniRocket模型处理。
请注意，本回答来自GPT-4
如果问题已经解决，请设置最佳答案