An Dijkstra Distance-based Clustering Algorithm And Application In Logistics

Recent years, the logistics industries have become the main pillar industries for economic development and main parts of economic growth in new century. Logistics distribution has great impact on profits and competence. At present, the entire work flow of logistics industry becomes more sophisticated. Because of rapid development of information, in order to control logistics activities effectively and allocate logistics resources reasonably, it is very necessary to use information technology to establish integrated and efficient logistics information decision solutions so as to reduce logistics costs.Operation research has wide range of applications in logistics management. In this paper, we used improved clustering algorithm to resolve practical problems existed in logistics management combined with operation research theory, of which are mathematical programming, graph theory and network, the theory of reasonable location, the theory of reliability.Clustering analysis uses in many fields widely, for example, market research, data analysis, pattern recognition, image processing, biology, marketing and telecommunications and so on. The applications of clustering analysis in different fields have different requirements. In this paper, we proposed an improved algorithm named DK-means based on Dijkstra distance according to the characteristics of logistics enterprises. The main works in this paper have several points as following: Firstly, we modeled the problems existed in logistics management as an undirected connected weighted graph. Secondly, we replaced the Euclid distance or Manhattan distance with Dijkstra distance. Thirdly, it can avoid the blindness of determining the value of k and randomness of initialing k cluster centers when we use the DK-means clustering algorithm to solve practical problems. Fourthly, we will divide the service scope of customers effectively if we use the improved algorithm in the procedure of logistics and distribution. Fifthly, we can construct minimum spanning tree for nodes which belong to the same sub-class to optimize the routes for distributing goods. Sixthly, according to particle size of division of space, we can apply the improved algorithm to all aspects of business process control management for logistics. After analysis combined with operations research, we can know it is very practical to apply improved clustering algorithm to logistics management.