Closed felix134 closed 6 years ago
Dear Felix,
thanks for your interest in ARX! This "issue" is not related to #54 but with the runtime complexity of the underlying optimal search algorithm (execution time grows exponentially with the number of attributes). Please see our publications for further details. To mitigate this, you can use the heuristic algorithm implemented in ARX. You can find an example, in Example34.java
Best Fabian
Thank you very much for your answer!
Hello. I've noticed that there is a significant performance drop in case of processing time, when I anonymize approximate 11 or more QIDs with the arx framework. The question is: Are there config values I can modify to recieve a better result? Or is this just an ordinary behavior of the used k-anonymity approach (with k=5)? Note: The problem I've encountered may be related to #54. (?)
Example: I tried to simplify my actual problem, and created tests structured as follows: I used for all tests a data set of size 10. For each test I removed a QID attribute (and its hierarchy). In the following an extract of the code with 11 attributes (10 of them are QIDs):
The results are: Number of QIDs | processing time
Sincerely, Felix Bölz.