First International Conference on MOLDAVIAN RISKS – FROM ...
First International Conference on MOLDAVIAN RISKS – FROM ...
First International Conference on MOLDAVIAN RISKS – FROM ...
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
<str<strong>on</strong>g>First</str<strong>on</strong>g> <str<strong>on</strong>g>Internati<strong>on</strong>al</str<strong>on</strong>g> <str<strong>on</strong>g>C<strong>on</strong>ference</str<strong>on</strong>g> <strong>on</strong> <strong>MOLDAVIAN</strong> <strong>RISKS</strong> - <strong>FROM</strong> GLOBAL TO LOCAL SCALE<br />
16-19 May 2012, Bacau, Romania<br />
DATA INCOHERENCE FOR COMBINATORIAL<br />
OPTIMIZATION PROBLEMS<br />
Gloria Cerasela Crisan 1 , Camelia-M. Pintea 2 , Camelia Chira 3<br />
1 “Vasile Alecsandri” University of Bacau, Department of Mathematics, Informatics and Educati<strong>on</strong><br />
Sciences<br />
2 Tech Univ Cluj-Napoca, North Univ Center Baia-Mare, Faculty of Science<br />
3 “Babeş-Bolyai” University of Cluj-Napoca, Department of Computer Science<br />
Corresp<strong>on</strong>ding author: Gloria Cerasela Crisan, ceraselacrisan@ub.ro<br />
Abstract: Data collecti<strong>on</strong>s are fundamental in taking correct and <strong>on</strong> time decisi<strong>on</strong>s. In risk<br />
management, the coherence and the integrity of the problem’s data have great impact <strong>on</strong><br />
the quality of the decisi<strong>on</strong>. If the data are biased, the applicati<strong>on</strong>s that assist the manager<br />
can provide poor advices. The real-life problems reflect the complexity and heterogeneity<br />
of our modern society. The applicati<strong>on</strong>s designed for solving them are therefore expensive,<br />
sophisticated and complex; inherently their results are hard to predict. This is why the<br />
influence of the data incoherence over the results is an extremely important issue to<br />
c<strong>on</strong>sider. The a-priori knowledge and its efficient handling are compulsory for modern,<br />
proactive strategy in problem solving. Knowing how a specific algorithm would behave<br />
when input data might be incoherent may help choosing between several algorithms athand<br />
to better deal with real-world problems. We here approach the influence of<br />
incoherent input data when solving Combinatorial Optimizati<strong>on</strong> Problems (COPs). Several<br />
new COP variants are defined and the stability of the corresp<strong>on</strong>ding solving methods under<br />
imperfect input data is investigated. Some parameters are defined and a parameterized<br />
complexity study is performed. Numerical experiments engage the well-known COP<br />
Travelling Salesman Problem in its standard form and a dynamic variant (with and without<br />
errors in the input data for both problem variants). All the problems are solved using Ant<br />
Col<strong>on</strong>y Optimizati<strong>on</strong> algorithms, a biologically-inspired metaheuristic approach. The<br />
adaptati<strong>on</strong> and the resilience of natural systems (Darwinian evoluti<strong>on</strong>, ant col<strong>on</strong>y<br />
behaviour) suggest the same behaviour. We use a natural incoherence measure and we<br />
compare the applicati<strong>on</strong>s results <strong>on</strong> correct and incoherent data. We modify the algorithms<br />
in order to store both the correct and the incoherent data. We extend our numerical<br />
experiments by coupling a new imperfecti<strong>on</strong> dimensi<strong>on</strong>: the uncertainty. The algorithms’<br />
behaviour does not have a pattern: we observe stable, unstable, even catastrophic effects of<br />
input data incoherence. The paper generalizes a metaheuristic approach to a classical<br />
optimizati<strong>on</strong> problem when the data entry has some degree of incoherence.<br />
Key words: combinatorial optimizati<strong>on</strong> problems, parameterized complexity, data incoherence,<br />
heuristics, ant col<strong>on</strong>y optimizati<strong>on</strong>.<br />
60