Research On Recommendation Technology And Application Implementation For Online Group Buying

With the developing and maturing of online payment technology and offline logistics network,the development momentum of online group buying has become stronger,and it is the next outlet of the e-commerce industry.However,the group purchase scene is complicat-ed,not only the purchase relationship between the buyer and the product,but also the group relationship between the buyer and other buyer,that is,multiple people buy a product togeth-er.In today's booming e-commerce,the problem of how to recommend suitable products to buyers has been deeply studied,while the research on how to recommend the right co-purchaser for a group initiator has not been studied.This article studies the recommendation technology in online group buying,and focuses on the issue of co-purchasers recommen-dation among them.In a group transaction,not only the strongly similar co-purchasers but also the weakly similar co-purchasers participate in the purchase,and these weakly similar co-purchasers must also be considered when recommending the co-purchasers.This article first proposes a connection-based co-purchaser recommendation scheme PDSDNE(PathSim Diffused Structural Deep Network Embedding).Group transaction data is processed into a trading network form.In the trading network,the weakly similar co-purchaser is explicitly connected with the initiating buyer,so that the initiating buyer is relat-ed to the weak similar buyer,Finally,The stack-type autoencoder is used to embed the trans-action network and obtain the embedding vector of the goods and buyer nodes.Based on the embedded vector,generate a recommendation list including weak similar co-purchasers.In order to further improve the recommendation performance,this paper proposes a co-purchaser recommendation scheme based on neighbors,named cop2vec(co-purchaser to vector),from another perspective of network embedding.cop2vec is a smoother network embedding recommendation scheme.It can obtain embedding vectors from the network which is beneficial to weak similar co-purchasers.There is no need to add additional infor-mation and it will not cause damage to the original network.The resulting embedding vector can generate co-purchaser recommendation list with higher accuracy and recall.In terms of system implementation,this article designs and implements a recommended application for online group purchase,which is developed using Django.The entire application is available,simple to operate,and has good visualization effects.Compared with the traditional recommendation methods,the recommendation methods proposed in this paper is more suitable for online group buying.When a buyer initiates a group purchase,this methods can effectively recommend other people,including weak similarity,to participate in group purchase,which will bring better shopping experience for the buyers on the group buying scenarios.