Page 35 - 33-3

NTU Management Review Vol. 33 No. 3 Dec. 2023

Table 11 Numerical Result of Number of Scale Level
Average Minimum
Number of scale levels z z GA z z GA
z* z* z* z*
1 0.9904 0.9202 0.6656 0.5964
3 0.9847 0.7952 0.7231 0.5189

Table 12 shows that the performance in three activity sessions is better. When there
are more activity sessions, the number of customers’ options also increases. Therefore,

even if the algorithm chooses non-optimal solutions, there is more chance that some of the
customers’ demand can still be served by other facilities.
Table 13 shows that the performance is better when the distribution of customer

location is cluster. It is because when customers are located clustered, it is likely to serve
more customers when building a new facility near the cluster center. As for the time
preference distribution, they perform better in random time preference distribution. If
customers tend to go to facilities in different activity sessions, there is less chance that
facilities are full, so it is easier for algorithms to get good solutions.

Table 12 Numerical Result of Number of Activity Sessions
Average Minimum
Number of activity GA GA
sessions z z z z
z* z* z* z*
1 0.9854 0.8566 0.7231 0.5189
3 0.9897 0.8588 0.6656 0.5848

Table 13 Numerical Result of Customer Location and Preference Distribution
Average Minimum
Location and preference z z GA z z GA
distributions
z* z* z* z*
cluster, cluster 0.9936 0.8568 0.7667 0.5848
cluster, random 0.9936 0.8561 0.8693 0.5755
random, cluster 0.9747 0.8558 0.6656 0.5189
random, random 0.9883 0.8622 0.7231 0.6471

27