E-commerce Applications Based On Multi-core Cluster Parallelization

The emergence of parallel computing brought epoch-making solutions to standalonecomputing era bottleneck, which also driven the parallel cluster development. Nowadays,with the development of computer clusters, parallel platform types gradually increased, eachplatform has unique advantages in parallel. Cloud computing is an Internet-basedsuper-computing model, which split distributed computing tasks on the computer cluster, getsuper-computing power, storage space and information services. It is the most popularcalculation mode.For the past few years, the e-commerce market from sellers market to buyers market,prompting a rat race between electricity suppliers. Customer relationship is the lodgment ofenterprise development, also the precondition of enterprise profit. Enterprises want to conductaccurate analysis of customers need to classify them. The traditional classification methodsare classified based on experience or a simple statistical methods. However, when facemassive data, stand-alone computing power is difficult. In this article, the parallel computingidea will be introduced into the e-commerce customer classification area to solve the aboveproblems. Designed a multi-table associated algorithm for data preprocessing, put commodityinformation and transaction history data obtained from the e-commerce site linked together,and converted into the form suitable for data mining. In addition, designed customerclassification methods, select FCM fuzzy clustering algorithm to analysis of pretreatmentcustomer data.The traditional way of data table association is using a local parallel database, but it ispowerless faced with massive data table association under multiple Internet. Hadoop clustercloud computing model can solve this problem. Hadoop cluster apply to large data-intensivecomputing tasks has high efficiency, and have applications in many areas. This paper basedon Hadoop cluster achieve a number of massive data tables associated and do detailedcomparisons of experimental data. The compared results showed that Hadoop cluster hasobvious parallel efficiency in handling massive data table association and large-scaledata-intensive computing tasks.After preprocessing of data analysis, select FCM algorithm to customer dataclassification, which is widely used in multivariate statistical analysis. Compared with thetraditional classification methods based on experience or simple statistics, it broadens theindex system, that is from a single index broadened to various indicators of customerspending patterns. Take Eslite transaction data as an example to test, according to customerconsumption patterns customers into four categories: quality customers, the general customers, small customers and potential customers. The experimental results verify the FCM algorithmfor data clustering effect and the high efficiency of MATLAB multicore parallel cluster inparallel complex algorithms processing.The approach designed in this paper can be applied to large-scale data processing in thefinancial sector and customer segment analysis, It has a certain application value.