The Doctor Rostering Problem - Asser Fahrenholz

Chapter 7. Tests, results and discussion 43
Table 7.1 displays the test instances and the problem factors that can affect the solution.
For every instance, Cons notes how constrained the problem is. I expect that adding
more night shifts, evening shifts and wishes (raising the Cons) also make the problem
harder to solve and likewise when decreasing the number of doctors.
Table 7.1: Test instances
Test Length (months) Cons (%) Doctors
0 4 13 6
1 4 17 6
2 4 12 7
3 4 13 5
4 8 13 6
5 4 18 6
6 2 17 7
7 6 13 8
8 4 5 6
9 4 0 6
The test instances can then be partitioned into the groups shown in table 7.2. This
shows the test instances that have equal values in what categories and will be used for
comparing results across groups in the next section.
Table 7.2: Test instance groups
Length Constrained Doctors
{0,1,2,3,5,8,9},
{4}, {6}, {7}
{0,3,4,7}, {1,6},
{2}, {5}, {8}, {9}
{0,1,4,5,8,9},
{2,6}, {3}, {7}
The problem factors will be the starting point for evaluating test results. From there, I
move on to discuss effects of the metaheuristic factors, such as the RCL parameter, the
enumeration and the weight of the constraints. The test instances have all been tested
with the following parameters:
RCL parameter {1 (Leaving only the best element in the list), 50 (cutting the list in
half), 100 (leave the list as it is)}
Enumeration Day, Shift
Constraint weights δi = 1, δi = 2
In the following, I denote the value of the objective function value Z(S) and the num-
ber of invalid hard constraints by V (S). Together they will be noted in the form of
(Z(S), V (S)).