Introduction

With the continuous development of e-commerce, e-commerce big data for the industry and academia has brought valuable opportunities and challenges [1]. E-commerce in bringing convenience to people’s life and work at the same time, more people is able to look forward to in-depth discovery and excavation of more valuable information and knowledge. Automatic identification of all web pages that describe the same commodity entity from a variety of electronic commerce data sources is the basis for data integration and data analysis. But the multi-source heterogeneous e-commerce data has a huge number of types of goods, different modes, uneven data quality, and diverse website structure characteristics of clutter. It lacks a unified model definition and theoretical model. These have greatly influenced the analysis and application of large data of electronic commerce. In this paper, a hybrid framework is proposed based on Map-Reduce architecture, which is based on data index, data integration, entity recognition and data ordering. In the big data environ ment the entity recognition is achieved by machine learning. Output to meet the same data sets and related attributes, and to sort. This research can be applied to the data fusion, data search and personalized recommendation in the electronic commerce environment.