基于Spark+MLlib分布式学习算法的研究

摘要

The key of E-commerce service is the usersˊ demands. With the rapid expansion of E-commerce business, and the enormously increasing of the usersˊ data amount, due to the fact that traditional single algorithm hardly meets the current business requirements, based on Spark+MLlib, the distributed learning algorithm is proposed, the system is able to carry on classification and prediction in the process of implementation, as well as the tag system. According to the testing, the new algorithm is obviously superior to single algorithm.%电子商务服务的关键是用户的需求，随着电子商务业务的急速扩展，用户数据量的海量增长，针对传统的单机算法很难满足业务需求的现状，提出了基于Spark+MLlib的分布式学习算法，系统在实现过程中进行了分类和预测，并实现了用户标签系统。通过测试，新的算法明显优于单机算法。