Monitoring e-commerce adoption from online data

The purpose of this paper is to propose an intelligent system to automatically monitor the firms' engagement in e-commerce by analyzing online data retrieved from their corporate websites. The design of the proposed system combines web content mining and scraping techniques with learning methods for Big Data. Corporate websites are scraped to extract more than 150 features related to the e-commerce adoption, such as the presence of some keywords or a private area. Then, these features are taken as input by a classification model that includes dimensionality reduction techniques. The system is evaluated with a data set consisting of 426 corporate websites of firms based in France and Spain. The system successfully classified most of the firms into those that adopted e-commerce and those that did not, reaching a classification accuracy of 90.6%. This demonstrates the feasibility of monitoring e-commerce adoption from online data. Moreover, the proposed system represents a cost-effective alternative to surveys as method for collecting e-commerce information from companies, and is capable of providing more frequent information than surveys and avoids the non-response errors. This is the first research work to design and evaluate an intelligent system to automatically detect e-commerce engagement from online data. This proposal opens up the opportunity to monitor e-commerce adoption at a large scale, with highly granular information that otherwise would require every firm to complete a survey. In addition, it makes it possible to track the evolution of this activity in real time, so that governments and institutions could make informed decisions earlier.