An enhanced soft set data reduction using decision partition order technique
Nowadays, redundant data is one of the open issues due to the rapid development of technologies. This issue is more visible especially in decision-making as the behaviour of such data is more complex and due to the uncertainty during a process of decision making. Besides, the need of extra memory is...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2017
|
| Subjects: | |
| Online Access: | http://umpir.ump.edu.my/id/eprint/23248/1/An%20enhanced%20soft%20set%20data%20reduction%20using%20decision%20partition%20order%20technique.wm.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Nowadays, redundant data is one of the open issues due to the rapid development of technologies. This issue is more visible especially in decision-making as the behaviour of such data is more complex and due to the uncertainty during a process of decision making. Besides, the need of extra memory is essential as redundant data makes use of storage and produce redundant copies due to its widespread use. Hence, the soft-set reduction techniques are introduced to assist in reducing storage space by facilitating less number of copies with minimum cost per line or per storage. The benefit of soft-set reduction is to foster the decision making process as well as to enhance the decision’s quality. Classification techniques that were previously proposed for eliminating inconsistency could not achieve an efficient soft-set reduction, which affects the obtained solutions; thus producing imprecise result. Furthermore, the decomposition based on previous algorithms could not achieve better parameter reduction in available domain space. The decomposition computational cost made during combination generation can cause machine infinite state as Nondeterministic Polynomial time (NP). The decomposition scenario in Rose’s and Kumar’s algorithms detects the reduction, but could not obtain the optimal decision. The contributions of this research are mainly focused on minimizing choices costs through adjusting the original classifications by decision partition order. Moreover, this research proposes a decision partition order technique to maintain the original classification consistency. The second contribution is enhancing the probability of search domain of Markov chain model. Furthermore, this research proposes an efficient Soft-Set Reduction accuracy based on Binary Particle Swarm optimized by Biogeography-Based Optimizer (SSR-BPSO-BBO) algorithm that can generate accurate decision for optimal and sub-optimal results. The results show that the decision partition order technique performs up to 50% in parameter reduction, while some algorithms could not obtain any reduction. On the other hand, the proposed Markov chain model could significantly represent the robustness of the proposed reduction technique in making the optimal decision and minimising the search domain by up to 33%. In terms of accuracy, the proposed SSR-BPSO-BBO algorithm outperforms other optimization algorithms by up to 100% in achieving high accuracy percentage of a given soft dataset. In addition, the proposed decision partition order technique has reduced the choices costs and thus improves the original classification consistency. Hence, the proposed technique could efficiently enhance the decision quality. Also, the accuracy of original soft-set optimal and sub-optimal results have been improved using an intelligent SSR-BPSO-BBO algorithm. The computational cost of search domain (space) has been enhanced using proposed Markov Chain Model. |
|---|