Based On Data Mining For Model Optimization Of Picking System

With the rapid development of China’s economy, especially in recent years, the rapid rise of e-commerce has a dramatic impact on the traditional logistics industry. Commodity supply of modern logistics is more for small species, diversity. Distribution center as a sign of modern logistics, it can reduce the number of transactions and circulation of products, produce benefits, reducing the number of customer inventory, ensure that inventory, it can also solve the challenges facing modern logistics. Based on the powerful functions and features of the distribution center, it also has become an important basis for the current e-commerce delivery logistics.Internal processes of distribution center including receiving, handling, storage, sorting, shipping and other operational procedures, and the order picking costs can account for 60% of the total costs of the distribution center, order processing time can account for 40% of the time of the distribution center, So if we can improve the operating efficiency of the order picking at no extra cost, it will have a positive impact on operational efficiency of distribution center. In this case, to product distribution policy items as a starting point, combined with the correlation between customer orders and items, we propose a demand-related items of cluster model in order to get the best items classification.From the perspective of theoretical analysis point of view, we propose one kind of items allocation strategy based on customer demand for small and medium sized logistics distribution center sorting system combined distribution center order demand, and adopts hierarchical clustering algorithm of direct dynamic clustering algorithm, put the items correlation coefficient as the clustering index, cluster all the items of the distribution center and get the final clustering results. For clustering algorithm for evaluation, using the Fisher distance algorithm, having the ratio of the variance of the distance between the clusters and cluster internal element as a reference and the average of the results obtained is seeking its validation parameters. Based on this algorithm, we need to use the concept of variance with the actual problem to analyze the distribution of items of clustering, and determine the best classification at last.The process of carrying out simulation experiments in this paper, through analysis of the order data and calculating the probability that the two items also appear in the same order to get the relationship matrix, and then deal with the relationship matrix using a mathematical formula to make it balanced and get away from the clustering index. Simulating the distance algorithm between clustering results by Matlab tool,validate and analyze the six different distance calculation method and combine with validation to get the best calculation method and cluster allocation strategy between items and analysis of the items allocation strategy to enhance the efficiency of the distribution center at the end.By assignment policy model for items set up, optimization, simulation, and analysis of their simulation results shows that a reasonable allocation policy items can have a good impact to the development of modern logistics and logistics quality services.