0

At first I implemented a SolverConfig.xml with a changeMoveSelector like this:

<changeMoveSelector>
            <entitySelector>
                <entityClass>Project</entityClass>
            </entitySelector>
            <valueSelector>
                <variableName>developer</variableName>
            </valueSelector>
</changeMoveSelector>

After 10 seconds of optimization I got the following result: -388hard/-65165soft

After that I uncommented the changeMoveSelector and after 10 seconds of optimization I got the following (better!) result: -350hard/-64713soft

How does OptaPlanner "moves", when there is no changeMoveSelector stated?

P. Krämer
  • 15
  • 5
  • 1
    See http://docs.jboss.org/optaplanner/release/6.2.0.Final/optaplanner-docs/html_single/index.html#moveAndNeighborhoodSelection, especially the section about `changeMoveSelector` – Ondrej Skopek Aug 18 '15 at 18:23
  • I read it again, but I do not see any hint, why I get a better result without stating a changeMoveSelector. – P. Krämer Aug 18 '15 at 19:48

1 Answers1

2

If no move selectors are defined, Local Search defaults to a union of change and swap moves - which explains your better result. It's in the docs IIRC.

Geoffrey De Smet
  • 26,223
  • 11
  • 73
  • 120