A fast clustering method for commodity-related big data sparse network

Abstract

The invention discloses a fast clustering method for commodity-associated big data sparse networks, comprising the following steps: preprocessing commodity sales record data, constructing and storing commodity-associated big data sparse networks; pruning commodity-associated big data sparse networks; Fast clustering is carried out on the sparse network of large data related to commodities, and the clustering results are obtained. The invention adopts a single-step linked list structure to store the relationship between commodity nodes, only marks the relationship of commodity nodes directly connected with the commodity nodes, and eliminates the redundant space in the common purchase relationship matrix. The invention prunes the low-degree commodity nodes of the commodity-associated big data sparse network to improve mining efficiency and clustering accuracy. The present invention can eliminate the constraints of traditional clustering algorithms by using the idea that high-value commodity nodes are divided by low-degree commodity nodes, and directly judge based on the degree value without considering data distribution, multiple access to data, and prior knowledge. Complete the clustering process.

Description

Technical field [0001] The invention belongs to big data processing technology, in particular to a sparse network processing method for commodity-related big data. Background technique [0002] With the comprehensive application of information systems and informatization construction, retail enterprises have produced a large number of data records in operation and management, and mining the operating viewpoints and market rules among these data has considerable application value. By studying the shopping transaction data of retail companies, mining the associations and frequent item sets between commodities, it can guide retail companies to make more scientific and reasonable category management, store layout display and product promotion and other operational decisions, and greatly improve the company’s Operating results. With the rapid development of e-commerce, e-commerce retail companies such as Amazon, JD.com, and Yihaodian provide high-quality product recommendation, relat...

AI Technical Summary

Problems solved by technology

The currently widely used adjacency matrix and adjacency linked list storage methods are used to analyze and mine potential laws and require multiple access to data, so they are not suitable for big data association analysis of retail products.

However, the clustering of large data sparse network oriented to retail goods needs to focus on improving efficiency. Commonly used complex network clustering needs to give an effective objective function and optimal solution search strategy in advance, and the search process sacrifices efficiency to improve the accuracy of clustering. , which is not suitable for clustering retail commodity association big data sparse networks

Images