Research On The Application Of The Artificial Fish Swarm Algorithm For The Clustering Problem

Clustering has been applied to many areas, including data mining, statistics, and machine learning and can be regarded as a global optimization problem. The Genetic Algorithm is a random search and optimized solution that mimics the process of natural evolution and genetics. Artificial fish swarm algorithm (AFSA) is a novel bio-inspired optimizing method. It is also a superincumbent design method, and has no particular demands on the form and property of the optimization space.The algorithm has strong self-adjustment ability and rapid convergence speed, also can overcome the local threshold and acquire the global best optimized value, and therefore could be applied into the calculation for many model optimizations. The artificial fish swarm algorithm provides a kind of novel method to solve the optimization problems.On the basis of advanced research on the artificial fish swarm algorithm and the genetic algorithm, this paper presents an algorithms combination of the artificial fish swarm algorithm and the genetic algorithm by means of combining the selection and mutation from the genetic algorithm with the artificial fish swarm algorithm.The research content and results are briefly described as below,1. On the basis of advanced research on the artificial fish swarm algorithm and clustering problems, this paper provides the artificial fish swarm algorithm on solving clustering problems which numbers of solution categories are known. The rough process is; firstly, to structure the individual artificial fish model, and then make proper connections between the individual artificial fish and the clustering problem, then confirm the objective function to the algorithm, lastly, to programme for the best solution.2. On the basis of advanced research on the solutions of the clustering problems and the artificial fish swarm algorithm, this paper presents the improved artificial fish swarm algorithm, which can avoid over rapid convergence speed at an earlier time and over slow convergence speed at a later time When the best optimized value has no continuous change or no obvious change, the improved artificial fish swarm algorithm involved in this paper applies variance and crossover operations in order to eliminate the artificial fish's aimlessly random movement or the massive convergences around the non-global threshold, and consequently gain the higher speed, better accuracy and improved quality in calculation. The experiment result is effective by chosen UCI data as the clustering data and has been emulated by the Java program.