A method for online retail sales estimation based on semantic features of web pages

Currently, e-commerce is being applied more and more widely in daily life. However, how to give an accurate, real-time and low-cost estimation of online retail sales is still a difficult problem from both academic and industrial aspects. This paper presents an efficient method for the estimation of online retail sales that is characterized by an order detection algorithm embedded in distributed clients to detect transaction amounts of successful orders. The proposed order detection algorithm is a kind of logistic regression classifier based on web semantic features, which divides web pages into three categories: ordinary pages, order placement pages and order confirmation pages. A further 10-fold validation is conducted and proves the algorithm is quite effective.