Dynamic Sale Prediction of the Time-Sensitive Products Based on Conformity

In these years, many online shopping and investment becomes more time-sensitive such as crowdfunding, online auction, time-limit sale and group buying. Most of them have deadlines. Consequently, people need to make faster decisions than other products leading to more conformity in this area. Though there is much work about the conformity and social influence in social networks, there lacks work of conformity in the online shopping and investment, because in these scenarios there is usually no such "strong" connection. However, we think the conformity information is also important for the sale prediction of time-sensitive products, since conformity impacts people's purchase decisions. In this paper, we propose a model which captures the dynamic conformity and apply it to the similarity computation and sale prediction of time-sensitive products. It can be dynamically adjusted to fit the target product and the prediction moment, which means it is well self-adaptive. Since it also gives the consideration to other static and latent features in a collaborative way, it alleviates the "cold start" problem nicely and is free from the presetting hypotheses. Experiments on the real large-scale dataset shows that our method exceeds other comparison methods by 33.2% in the sale prediction.