PEPP: profits estimation in prices promotion

An effective and efficient algorithm was proposed to estimate the profits of the sales in prices promotion for retailing and other business applications based on the historical transactions data. It predicts the changing of each customer's purchase under the specific prices promotion by estimating the effects from the promoted items. The distance from the promoted items to the items before promotion is used to estimate the effects. The approach considers both the increase of customers' number because of promotion and the cross-selling effects between items. It models customers' purchase actions well and is suitable for real business applications. Experiments showed that it is efficient and effective and can predict well the sales in prices promotions and is also highly scalable to the data set with large number of items and transactions.