Based On The Co - Supermarket Distribution Analysis Of The Location Pattern Mining

With modern social economic development and the rhythm of urbanization rhythm faster and faster, the competition of market environment and social economic environment more and more fierce, reasonable facility location decision will decided to the fate of the business, and have a directly affect to its fate. For the owner, through the analysis of location problem, can determine the most investment potential shop site。 In recent years, spatial data mining has been widely studied and applications, it in solving the space knowledge acquisition has lot of progress. That spatial data mining technology uses in studying the distribution laws of the supermarkets in the city, can provide feasible ideas and the reference for the reasonable choice of large supermarket’s address.This paper first proposed a co-location mining algorithm based on k-means clustering, the idea of algorithm based on the inner relationship of clustering and space co-location model, the k-means clustering algorithm is introduced to space co-location model mining. Firstly, pick up the data of supermarkets and other urban spatial elements relevant with supermarkets, then use k-means clustering algorithm respectively to each space object that need to be mined for forming clusters, defined the concept of cluster proximity and coarse table instance, calculating the coarse table instance and rough pruning,then calculate real instance of the table that have not been pruning. Mining co-location model in such a way, as k-means clustering algorithm is simple、arameter K is easy to understand and other advantages, make the space co-location model mining application more realistic.Then, this paper takes Kunming for example, chose Carrefour、Wal-Mart and another six urban space objects as the research objects, through the use of k-means-based co-location model mining algorithm, and setting distance threshold and prevalence threshold parameters with different values, mining co-location rules in different conditions. Then analyzing mining results, the distribution laws of big supermarkets are obtained. To a certain extent, our work would provide feasible methods for the supermarket decision makers, when they make decisions.