Convert mat file to pandas dataframe

Question

I want to convert this file to pandas dataframe.

import pandas as pd
import scipy.io
mat = scipy.io.loadmat('cardio.mat')
cardio_df = pd.DataFrame(mat)

I get this error : Exception: Data must be 1-dimensional

Brogan

Answer 1

It seems mat is a dict containing X of shape (1831, 21), y with shape (1831, 1), and some metadata. Assuming X is the data and y are the labels for the same, you can stack them horizontally with np.hstack and load them into pandas:

In [1755]: mat = scipy.io.loadmat('cardio.mat')

In [1758]: cardio_df = pd.DataFrame(np.hstack((mat['X'], mat['y'])))

In [1759]: cardio_df.head()
Out[1759]: 
         0         1         2         3         4         5         6   \
0  0.004912  0.693191 -0.203640  0.595322  0.353190 -0.061401 -0.278295   
1  0.110729 -0.079903 -0.203640  1.268942  0.396246 -0.061401 -0.278295   
2  0.216546 -0.272445 -0.203640  1.050988  0.148753 -0.061401 -0.278295   
3  0.004912  0.727346 -0.203640  1.212171 -0.683598 -0.061401 -0.278295   
4 -0.100905  0.363595  1.321366  1.027120  0.141359 -0.061401 -0.278295   

In [1760]: cardio_df.shape
Out[1760]: (1831, 22)

Duwe

@COLDSPEED How about converting to an array-like with shape = (n_samples, n_features) ? – Brogan

@KouchakYazdi Do you mean, you want to reshape it first? – Duwe

@COLDSPEED What if I want mat file in an array-like with shape = (n_samples, n_features) format ? (not a dataframe) – Brogan

@KouchakYazdi data = df.values.reshape((n_samples, n_features)) – Duwe