|
Partial optimisation example |
  
|
|
Partial optimisation example |
|
We now wish to demonstrate the practical application of partial optimisation.
First remove all existing course assignments , delete all existing clusters and then launch partial optimisation .
Since there are no clusters, all course weekly periods are still to be scheduled, and the course list therefore contains all the courses of our school.
In the first step we are going to schedule all the courses with 5 periods/week. Start by sorting the course list according to the number of periods by clicking on the "Per" column heading. Then hold the left mouse-button move the cursor across all the 5-period courses. Now press <SPACE> or click in the 'Sel.' column to mark the courses.
We will first try to manage with three clusters, and so we enter a '3' in the field 'Number of clusters to be created'.
Click on the <Continue> button to proceed to the next step. No problems are encountered and no warnings are displayed, and so the ' Opt. of subsets ' window is displayed. We do not wish to set any constraints/parameters so we can launch optimisation immediately by pressing the < Optimisation >.
Untis finds a solution without clashes almost immediately and displays the message 'Optimisation completed - Solution found'. Confirm this with<\<>OK>. You can see how courses have been scheduled in clusters in the course-cluster-matrix .
Now press <OK> in the "Opt. of subsets" window and confirm the save cluster prompt with <YES>. The 'Definition of the subsets' window will be redisplayed.
In the next step, we are going to schedule all courses with 3 periods/week and see whether we will be able to manage with just 2 clusters. Mark all the 3-period courses, enter 2 in the "Number of clusters to be created" field and then press <Continue>. A message is displayed informing us that student Talisker has too many courses in our selected subset, i.e. 3, but we only allow the creation of 2 clusters. We could now press <Cancel> and change our input, but we would first like to know whether there are any other students with too many courses before we decide on a new number of clusters. The next message window lists a whole number of students with too many courses. We now press <Cancel> and enter 3 as the desired number of clusters to be created. Clicking on <Continue> and confirming the message with <OK> takes us to the optimisation dialogue.
Start optimisation again, and almost instantly there is a solution Click on <OK> to once more save the clusters.
In the next step we are going to schedule the 4-period clusters together with the 2-period clusters. Please mark all the relevant courses with and test if 4 clusters are sufficient. After clicking on <Continue> we see the message that one student has 6 courses in the subset. We therefore increase the number of clusters to 6 and reach the optimisation dialogue without any further messages.
After optimisation is launched, a solution is quickly found but Untis is not completely satisfied with it and continues the calculation. After a certain time a solution is found that does not violate any boundary conditions, and the message 'Optimisation completed, solution found' is displayed. We save the clusters once more and now see that only the two remaining weekly periods of the German courses have not been scheduled.
We could now have partial optimisation schedule the remaining German lessons in a new cluster. However, we decide on scheduling using standard optimisation since these remaining course periods can perhaps be added to existing clusters. The clusters from the partial optimisation are already fixed meaning that they can no longer be modified, and we can call standard optimisation immediately. In the standard optimisation dialogue we check the box ' Add courses to locked clusters ' and launch optimisation.
As you can see, standard optimisation does not succeed in scheduling the remaining German periods in existing clusters. We nevertheless obtain a significantly improved result compared to the previous standard optimisation with a total of 38 cluster periods.
If you invest knowledge of the course structure at your own school in defining subsets, partial optimisation will generally reward your effort with results that are far better than those of standard optimisation.