Question

what happens when the fitness function in a genetic algorithm is giving worse solutions a better fitness value than solutions that are better in the search space? Should I try using different fitness functions or multiple fitness functions?

Answer 1

As we know, the choice of a good fitness function is very important in order to guide the GA towards the optimum solution.

But there is no fitness function suitable for all types of problems. Every problem uses it's very own fitness function, either a pre-defined fitness function or human-designed fitness function.

Error based measures and entropy-based measure are often used as fitness functions in classification problems.

An unrefined model of a given function (under investigation) is used for calculating fitness in optimization problems.

Dynamic Fitness functions are used quite often. A dynamic fitness function changes during the evaluation to explore search space in a better fashion.

So you should choose a fitness function which is suitable for your problem type to start with.

Let me know the type of problem you are dealing with in comment section.