A Case-Based Heuristic Selection System:
A Case-Based Reasoning heuristic
selection system for university timetabling has been developed and implemented.
Two components are included:
- Heuristic selection from 'constructive methods' (e.g. graph heuristics) and
- Heuristic selection from 'meta-heuristics' (e.g. tabu search, simulated annealing, variable neighbourhood search).
Two case bases (storing cases with graph heuristics and meta-heuristics) have been built (see Fig.1).
Figure 1: Case Based Heuristic Selection - System Interface
The system has the capability to learn what works well (see Fig. 2). When a new timetabling
problem is tackled, the corresponding case can be added into the case base after a
training process. For case-based heuristic selection on 'constructive methods',
the new problem will be solved by employing a hyper-heuristic which chooses from graph
colouring heuristics. For case-based heuristic selection on 'meta-heuristics',
the new problem is added into the case base with the best method obtained.
Figure 2: Case-Based Heuristic Selection - Training on New Cases
For further details about case based reasoning please visit this page.