Information Technology Reference
In-Depth Information
Fig. 2. The Average Flow Time of 200 Executions
Fig. 3. The Average Flow Time and Throughput of 100 Cases
about 0.8, and the result with SR is about 0.9. The throughput with SR is a little
better. The result means workflow management system can finish more cases in
unit time if considering previous resources influence on candidate resources.
7 Conclusions
In this paper, we introduce a method to capture the social relation between two
resources, and then compute the influence of the previous resources on the can-
didate resources according to the social relation. The experiment was conducted
with a real life log and a simulated data set. The results show that consider-
ing the social relation for task allocation will improve the performance of the
workflow management system. And the Q-learning algorithm is proved to be
suitable for task allocation. So it is necessary to consider the social relation for
task allocation. However, the experiment considers little about workload. So the
resources workload should be defined more accurate in the future work.
Acknowledgment. This work is Supported by the National Natural Science
Foundation of China under Grant No.60873162; the Natural Science Foundation
Search WWH ::




Custom Search