How to estimate the number of test cases which have to be present in "Non Regression Pack"?

I have to estimate the number of test cases which have to be present in "Non Regression Pack" for the next iteration.

To do that I have different data:

• Topic name
• Number of test case for each topic
• Business Impact by topic (High, Medium, Low) gave by business people
• Usage in production by topic (High, Medium, Low) gave by business people
• History of the previous iteration like the number of defect by topic (some topic are the same)
• Number of man days for each topic

What would be your approach to estimate the number of regression test case by topic?

Thank you for help and advice.

Best regards