nuclearstar / k-anonymity Goto Github PK

Hiya! I just came across this looking for k-anonymity implementations, I saw that you used our notebook from the EuroPython 2018 (https://github.com/KIProtect/data-privacy-for-data-scientists). It would be nice if you could add a backlink to it and put the proper license in your code (our code was licensed under MIT). Thank you! If you have questions regarding the implementation I'm happy to help.

AttributeError Traceback (most recent call last)
in
----> 1 dfn = build_anonymized_dataset(df, finished_partitions, feature_columns, sensitive_column)

in build_anonymized_dataset(df, partitions, feature_columns, sensitive_column, max_partitions)
14 grouped_columns = df.loc[partition].agg(aggregations, squeeze=False)
15 sensitive_counts = df.loc[partition].groupby(sensitive_column).agg({sensitive_column : 'count'})
---> 16 values = grouped_columns.iloc[0].to_dict()
17 for sensitive_value, count in sensitive_counts[sensitive_column].items():
18 if count == 0:

AttributeError: 'list' object has no attribute 'to_dict'

I tried converting this python code to pyspark code. I am running the same dataset with pyspark code in AWS EMR cluster.
For 200 records it was taking 9 minutes of time. For the 30,000 records it was taking 22.5 hours of time. Is there any way to optimise the code? Please help me.
Thanks in Advance.

replace line 173 with : values = {'age' : grouped_columns[0], 'education-num' : grouped_columns[0]}
now we no longer have the error list object has no attribute 'to_dict()'

then while printing your base you get the error " unhashable type: 'list' "
you just need to replace all the prints of your k-anonyme,l-diverse and t-close databases that are named dfn , dfl and dft by print(dfn.head())...

then the code works perfectly