CODEWITHSUNDEEP
pythonpandas
Pelide
Pelide

Reputation: 528

How to replace True/False with value on if condition?

I'm comparing the (sub)string of two columns from a data frame.

By following the suggestions from this thread I'm able now to put the condition TRUE if there is a correspondence between the column manual_label and prediction and FALSE if there is no correspondence.

Here a snapshot of the code I used:

argument_cols = ['prediction']
boolean_idx = df[argument_cols].apply(
    lambda arg_column: df['manual_label'].combine(arg_column, lambda token, arg: token in arg)
)

df['boolean_idx'] = boolean_idx 
pd.options.display.width=None
print(df)

df.to_csv('csv_file_w_pred.csv',sep=',',index=False)

The resulting data frame looks like this:

subject  manual_label                                prediction                                    value   boolean_idx
A  __label__Religione_e_Magia                  (__label__Bibbia_storie_dal_Vecchio_e_dal_Nuov...  ...        False
B  __label__Religione_e_Magia                  (__label__Religione_e_Magia,__label__Storia)       ... True
C  __label__Mitologia_classica_e_storia_antica (__label__Societa_civilizzazione_cultura,)         ...        False
D  __label__Essere_umano_uomo_in_generale      (__label__Essere_umano_uomo_in_generale,)          ...        True
E  __label__Religione_e_Magia                  (__label__Religione_e_Magia,)                      ...        True

The column prediction can have multiple labels.

However, if the condition is True I would like to have the value that satisfies the condition and if the condition is False I would like to have the 1st value from the col prediction which, as I said before can have multiple labels.

Desired output:

subject  manual_label                                prediction                                    value   boolean_idx
A  __label__Religione_e_Magia                  (__label__Bibbia_storie_dal_Vecchio_e_dal_Nuov...  ...   __label__Bibbia_storie_dal_Vecchio_e_dal_Nuovo_Testamento
B  __label__Religione_e_Magia                  (__label__Religione_e_Magia,__label__Storia)       ...   __label__Religione_e_Magia
C  __label__Mitologia_classica_e_storia_antica (__label__Societa_civilizzazione_cultura,)         ...   __label__Societa_civilizzazione_cultura
D  __label__Essere_umano_uomo_in_generale      (__label__Essere_umano_uomo_in_generale,)          ...   __label__Essere_umano_uomo_in_generale
E  __label__Religione_e_Magia                  (__label__Religione_e_Magia,)                      ...   __label__Religione_e_Magia

Suggestions?

Regards

Upvotes: 1

Views: 567

Answers (1)

Pelide
Pelide

Reputation: 528

If anyone needs I solved the issues as follows

argument_cols = ['label_1','label_2','label_3']

boolean_idx = df[argument_cols].apply(
    lambda arg_column: df['manual_label'].combine(arg_column, lambda token, arg: token in arg)
)

selected_vals = df[argument_cols][boolean_idx]
selected_vals = selected_vals.replace(np.nan, '', regex=True)
selected_vals = selected_vals.applymap(str)
df['suggested_label'] = selected_vals["label_1"].astype(str) + selected_vals["label_2"]+ selected_vals["label_3"]

df = df.replace(r'^\s*$', np.nan, regex=True)
df.loc[df['suggested_label'].isnull(),'suggested_label'] = df['label_1']
print(df)

Upvotes: 1

Related Questions