In this paper, we present a new genetic algorithm with floating point alphabet based on the ideas of algorithm proposed by Alp et al.
GA with greedy heuristic and floating point alphabet.
[3, 42] (Algorithm 6), its version for initial cluster centers seeding only (Algorithm 7), our modifications allowing elimination of many cluster centers in one step (Algorithm 8) and our new genetic algorithm with floating point alphabet (Algorithm 11).
In this case, for all solved large-scale test problems with both Euclidean ([l.sup.2], planar p-median problem) and squared Euclidean ([l.sup.2.sub.2], k-means problem) metrics, our Algorithm 10 with floating point alphabet and modified greedy crossover heuristic (Algorithm 11) works faster and gives more precise results than Algorithm 5 with greedy heuristic implemented for the initial seeding only (Algorithm 7, [3, 42]).
Diagrams show that new algorithm with floating point alphabet allows to increase the accuracy at early stages of the computation process in comparison with known methods which allows to use it for obtaining quick approximate solutions.