Research On Improved Artificial Bee Colony Clustering Algorithm

Recent years, the rapid development of computer network and the intensifying degree of market competition have brought greater challenges to efficiency of clustering solution in customer relationship management problems. Artificial bee colony algorithm has received the attention of scholars at home and abroad, and applied in engineering successfully due to its simpler and more effective robust characteristic than other heuristic bionic algorithms. Artificial bee colony algorithm has been proved in practice to be of quick and effective solubility. However, design of artificial bee colony algorithm is too dependent on the characteristics of problems. A design of algorithm design recommendation mechanism automatically based on the characteristics of the problem can solve the problem of low efficiency and repeatability of the problems of traditional algorithm design methods and achieve a rapid design of problem driven algorithm. This paper establishes an improved artificial bee colony algorithm which effectively improves the efficiency of clustering based on generally abstraction for problem and algorithm. In this paper, the main research work and innovation points include:(1) The establishment of a general description model of clustering analysis algorithmThrough the analysis of the object of customer relationship problems and its properties, the characteristics of the problem of attribute clustering are sorted out and then clustering description model of customer relationship problem is established based on ontology theory.(2) The establishment of improved artificial bee colony algorithm engineThis paper carries out a comprehensive review around the principles, characteristics, colony algorithm improvement and its application and other aspects, focusing on the study of bee colony algorithm for complex environment and its multiple improved forms and analyses the future research directions. In order to enhance the development ability of exploration and integrate the global optimal solution into the search process, an improved artificial bee colony algorithm is proposed. Improved artificial bee colony algorithm effectively improves the performance of artificial bee colony algorithm. Compared with the contrast method, the improved artificial bee colony algorithm has higher convergence precision, and faster convergence speed.(3) The establishment of a hybrid algorithm for solving frameworkProblem-strategy knowledge base is constructed based on the problem of customer relationship clustering analysis and artificial bee colony algorithm engine. Colony operator selection mechanism based on probability is established by recording operator choose history performance of each kind of question and iteratively differentiating Selection probability of different operator combinations to achieve a final recommended strategy for operator. In the process of the establishment of the mechanism, this thesis designs two kinds of method to generate the initial operator pool:one is the method based on artificial bee colony algorithm complete design rules, the other is regard the operator design rule sample obtained from the method based on uniformed clustering design as the initial operator pool. Advantages of regulating the ability of global optimization and local optimization of the improved artificial bee colony algorithm combine with the advantage of Fast convergence speed of K-means algorithm improve the robustness of the algorithm. Through many experiments and analysis, this algorithm not only has overcome the disadvantage of Poor stability of the traditional K-clustering algorithm, but also obviously improved the clustering effect. It is also proved that the two methods are of rationality and validity.(4) The framework instantiation and method studyAccording to the optimization strategies, instantiation is aimed at the problem of the customer relationship management in the process of instantiating framework. Through problem domain modeling, rule base realization and method study, it succeeds in realizing the improved colony algorithm engine in detail. Through the quantitative description of the classifier and research on recommendation algorithm method, operator framework of this kind of problems is instantiated. A large number of experiments and comparative analysis have proved that the hybrid algorithm is feasible and rationality.