Consumer Type Identification Research In The Context Of E-commerce

With the development of e-commerce,there are more and more opportunities for consumers to choose,market competition is becoming increasingly fierce,consumer instability is increasing,and the main competition between enterprises is gradually transformed into competition for customer resources,which requires enterprises to be able to efficiently understand the needs of consumers,so as to obtain a competitive advantage in the industry.In order to improve consumer satisfaction,it is particularly important to provide consumers with personalized services that meet their needs.However,due to the complexity and diversity of consumer preferences,merchants cannot meet all consumer needs,e-commerce operators want to achieve efficient customer relationship management,must identify user types according to the personalized characteristics of users,target different user groups,adopt corresponding marketing strategies,so as to improve user stickiness,reduce enterprise operating costs,enhance their competitiveness and profitability.With the maturity of information technology,enterprises can use big data analysis technology to achieve accurate classification of consumers,thereby improving the decision-making level of enterprises.In view of the above problems,based on the complex network theory and complex network community structure division algorithm,this paper is studied from the following three aspects:Firstly,from the perspective of B2 C e-commerce network composition,the evolution drivers of B2 C e-commerce network network feature network are analyzed,and on this basis,combined with relevant theoretical research,the evolution rules of e-commerce network are analyzed.According to the rules of B2 C e-commerce transactions,two improvements are proposed to the node-based strength evolution model,(1)Due to the limited vision of consumers and the existence of the local world,the consumer merit-based mechanism is no longer for the entire network,but for the local world where each node is located,and consumers choose the connection node in the local world,which can effectively reduce the cost of information search.(2)The evolution model based on node strength is only a growth model,and in the actual e-commerce network,there will be nodes joining and exiting nodes,so this paper adds the old node demise mechanism.Based on the above two points,the network evolution model based on node strength is improved,and a local optimal evolution model based on node strength is constructed,and the complex network characteristics of B2 C e-commerce network with small world and no scale are obtained by the average field method.Secondly,due to the different contributions of different goods to the similarity of consumers,this paper uses IEM as a community structure division algorithm to divide consumers into community structure,and proposes two improvement measures for the part of IEM algorithm that is not suitable for e-commerce networks,(1)Aiming at the phenomenon that the original algorithm randomly selects the initial node and causes the unstable division of the community community,this paper proposes a new node influence evaluation index IMC,which determines the initial cluster center according to the size of the node influence value.(2)When the original algorithm expands the community,it is necessary to know the overall structure of the network with the goal of optimal modularity,the complexity of the algorithm is high,and the original algorithm overemphasizes the connection strength between communities and ignores the similarity between communities in the process of community merger.This paper proposes a community structure discovery method based on local expansion to gradually scale outward from a given node,which overcomes the shortcomings of the global algorithm and enhances the accuracy of community division results.Finally,in order to accurately identify the arbitrarily shaped cluster structure in the e-commerce network,this paper applies the density-based clustering algorithm to the e-commerce network.Aiming at the problem that arbitrary selection of clustering parameters in DBSCAN clustering algorithm affects clustering accuracy,a BDSCAN clustering algorithm based on improved particle swarm optimization is proposed,which first proposes an inertia weight update strategy with adaptive particle variance to solve the phenomena of local optimum,premature convergence and weak search ability caused by the loss of population diversity in the process of particle swarm optimization.Then,the modularity function of evaluating the clustering effect is used as the fitness function of the algorithm,and the clustering parameters are selected accordingly,and the improved algorithm and the traditional algorithm are compared and tested on the dataset,which proves the stability and accuracy of the algorithm.The algorithm is applied to the e-commerce network dataset to verify the effectiveness of the algorithm.