CODEWITHSUNDEEP
pythonpandasnumpynan
k.ko3n
k.ko3n

Reputation: 954

Adding a nan values to dataframe gives unhashable type: 'numpy.ndarray' in Python

I want to insert np.nan to a dataframe; one nan for each row in random positions.
This is my dataframe: 

list_cols= ['col01', 'col02', 'col03', 'col04', 'col05','col06', 'col07', 'col08', 'col09', 'col10','col11', 'col12', 'col13', 'col14', 'col15', 'col16']

X_full = pd.DataFrame(np.random.uniform(low=1.0, high=100.0, size=(5,16)), columns=list(list_cols))

This is my code: 

# Add a single nan value to each row
rng = np.random.RandomState(0)
n_samples, n_features = X_full.shape
X_missing = X_full.copy()
missing_samples = np.arange(n_samples)
missing_features = rng.choice(n_features, n_samples, replace=True)
X_missing[missing_samples, missing_features] = np.nan

It returns TypeError: unhashable type: 'numpy.ndarray'.
Thanks for help.

Upvotes: 0

Views: 180

Answers (2)

Ayoub ZAROU
Ayoub ZAROU

Reputation: 2417

you could do 

X_missing = X_full.copy()
indexes = np.random.choice(range(X_missing.shape[1]), X_missing.shape[0])
X_missing.values[range(X_missing.shape[0]), indexes] = np.nan

Upvotes: 1

Arno Maeckelberghe
Arno Maeckelberghe

Reputation: 375

Since I'm not 100% sure I understand your question correctly, in case you just want to change the value of a single cell into NaN (i.e. values at (0, 12) & (1, 7) etc should be nan) then you can use:

for row, column in zip(missing_samples, missing_features):
    X_missing.iat[row, column] = np.nan

Note that we use a method starting with an 'i': .iat. This means the change we are applying is based on an Index and not on a row/column name.

In your example you get an error because it interprets missing_samples and missing_features as row names and column names, but the values you gave were number (namely the indices). You could use .iloc to clarify that what you are giving is an index and not a name but then it will be replacing whole rows, that's why I use .iat because I assume you only want to replace values 'at' specific 'i ndices'. Hope this helps

Upvotes: 1

Related Questions