Study On The Algorithms For The Facility Location Problem Based On Cluster Analysis

The main content of this article is facility location problem. The traditional point to point transportation routes was transformed into the continuous circle routes in the un-capacitated facility location problem which is closer to the actual type transport routes, making the model more practical. The discrete particle swarm optimization was applied into K-Harmonic means clustering algorithm as the calculation method, and was used for solving the model, experimental results proved its superiority.The first part of the paper described the significance of studying the facility location problem and logistics play an important role in the modern society; through summarizing the work done previously and the main problems currently, proposed the idea of calculating the facility location model using cluster analysis and applying the discrete particle swarm optimization into cluster analysis. The second part reviewed the origin, development process and the present situation of the facility location problem. The concept of logistics centers, functions as well as the various popular facility location model was introduced, and the features and algorithms of these models was summarized; the un-capacitated single-stage facility location model used through this paper was highlighted and the calculation method of the facility location model was summed up. The third part outlined the cluster analysis, described from the origin, developing of data mining to the current state of development in detail; clustering analysis as the important branch of data mining was focused on. The definition of cluster analysis, measure distance between classes and in class was illustrated. Two clustering methods:hierarchical clustering and K-Harmonic means clustering method was compared, and their strengths and weaknesses and scope of application were discussed.The main content of this paper are the fourth and the fifth parts. Through the introduction of the facility location problem and the cluster analysis, trying to combine them by using cluster analysis to study the facility location problem. Clustering the point set to several subset, each subset was supplied with products from a logistics center, and then connect every subset by the method of the solution of the Chinese postman problem, then obtained the transport costs of every subset. So the traditional point to point transportation routes was transformed into the continuous circle transportation routes which is closer to the reality, and numerical experiments show the superiority of this thinking.Part V described the origin and theory of the particle swarm optimization, and the difference and connections of the continuous and discrete versions of the particle swarm optimization algorithms. The discrete particle swarm was introduced into cluster analysis. combined with the strong local search feature of the K-Harmonic means clustering method and the global search feature of the particle swarm was combined to get an improved K-Harmonic means clustering algorithm to faster close to the optimal solution. Experimental results proved its effectiveness.Finally, all the work done in the paper was summarized, and the prospect work of the future was given.