Fix agg_numerical_column function

[ThaminduR]

Logical error in the agg_numerical_column has been fixed.

Related Issue - #13

[agessner]

Hey @ThaminduR thanks for fixing it! I don't know much about k-anonymization, but here are the tests @Marcaoandradenogueira and I have made. They are giving us the following results:

input

column1	column2	column3	column4	column5
600000	1	test1	x	20
600000	1	test1	y	30
800000	2	test2	x	50
800000	2	test3	x	45
800000	1	test2	y	35
400000	2	test3	y	20

params

categorical = set(('column2', 'column3', 'column4'))
sensitive_column = 'column4'
feature_columns = ['column1', 'column2', 'column3']
k = 2

k_anonymize

column1	column2	column3	column4	count
400000-600005	1, 2	test3, test1	x	1
400000-600005	1, 2	test3, test1	y	2
800000	1, 2	test3, test2	x	2
800000	1, 2	test3, test2	y	1

k_anonymize_w_user

column1	column2	column3	column4	column5
05-10	1, 2	test1, test3	x	20
05-10	1, 2	test1, test3	y	30
05-10	1, 2	test1, test3	x	20
800000	2, 1	test2, test3	x	50
800000	2, 1	test2, test3	y	45
800000	2, 1	test2, test3	y	35

The k_anonymize appears to be fixed, but the k_anonymize_w_user is having the same problem as before. Maybe it's some parameter we are passing wrong or the data is not a good fit. It just felt weird having this 05-10 result.

Thanks!

[ThaminduR]

97c2211 should fix the issue with k_anonymize_w_user function.
Also ,now the numerical aggregating function will handle 0 or 5 as the last digit case, gracefully.

[ThaminduR]

if the test cases are giving the correct output, then I'll merge this.