| Algorithm 2: Credibility and coverage based rule extraction algorithm |
Input: List (d-table), threshold of credibility: α, threshold of coverage: β | |
Output: List (rule) | |
1 | Traverse every item in List (d-table) as dt |
2 | Traverse every equivalence class in dt as ec-d |
3 | Calculate the credibility of ec-d, named as u |
4 | Calculate the coverage of ec-d, named as c |
5 | If u <α or c < β, remove ec-d from dt |
6 | Traverse every item in List (d-table) as dt |
7 | Traverse every equivalence class in dt as ec-d |
8 | Generate a series of rules R in every possible combination of condition attributes of ec-d |
9 | Traverse every item in R as r |
10 | If r doesn’t conflict with other rules and r has the less number of condition attributes, remove the rules with the more number of condition attributes from the same ec-d |
11 | Return the final R, List (rule) |