在CSV文件中编写和读取浮点数和字符串-python

Question

I am a bit new to python and programming. In my code, I have developed a feature (which is a 1-D array of 39 elements) for each audio file. I want to write the name of the file, the feature and its target value {0,1} into a CSV file to train my SVM classifier. I used the CSV writer as follows.

with open('train.csv', 'a') as csvfile:
    albumwriter = csv.writer(csvfile, delimiter=' ')
    albumwriter.writerow(['1.03 I Want To Hold Your Hand'] + Final_feature + [0] )

I want to write the details of around 180 audio files to this CSV file and feed it to the SVM classifier. The code that I use to read the file is:

with open('train.csv', 'rb') as csvfile:
    albumreader = csv.reader(csvfile, delimiter=' ')
    data = list()
    for row in albumreader:
        data.append(row[0:]) 
data = np.array(data)

I can access the name of the file in the first row as data[0][1] and the feature as data[0][2] but both of them are in <type 'numpy.string_'> . I want to convert the feature into a list of floats. The main problem seems to be the ',' that separates the elements in the list. I tried using .astype(np.float) but in vain.

Can anyone suggest me a good method to convert the strings from the CSV file back to the floats? Your help is very much appreciated as I have very less time to complete this project. Thanks in advance.

Edit: As per the comment, this is how my train.csv looks like:

"1.01 I saw her standing there" "[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38]" 0

"1.02 I saw her" "[2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40]" 0

"1.03 I want to hold your hand" "[3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41]" 1

Answer 1

I don't get exactly what you want to achieve, but assuming Final_feature is a python list of floats, and according to your code snippets for writing the csv file, you get the list as a string which probably looks like this: (which you get in data[0][2])

feature = '[3.14, 2.12, 4.5]' # 3 elements only for clarity

You asked how to convert this string to float, you can use:

map(float, feature[1:-1].split(','))

For reference, map applies its first argument to every element of its second argument, thus transforming every string in a float and returning a list of floats.

Another solution would be to write each element of your Final_feature in a separate column.

Answer 2

To convert string like "[1.0, 2.0, 3.0]" to list [1.0, 2.0, 3.0]:

# string to convert
s = '[1.0, 2.0, 3.0]'

lst = [float(x) for x in s[1: -1].split(',')]

# and result will be
[1.0, 2.0, 3.0]

This works both with standard python string type and with numpy.string type.

Answer 3

From what I can see, the variable Final_feature is a list of floats? In which case based on how you wrote the file the following will import the data

with open('train.csv', 'rb') as csvfile:
    albumreader = csv.reader(csvfile, delimiter=' ')

    audio_file_names = []
    final_features = []
    target_values = []
    for row in albumreader:
        audio_file_names.append(row[0])
        final_features.append([float(s) for s in row[1:-1]])
        target_values.append([int(s) for s in row[-1]])

There are two list comprehensions to convert the data into floats and integers.