Product Adoption Rate Prediction: A Multi-factor View

Abstract

As the worlds of commerce and Internet technology become more inextricably linked, a large number of user consumption series become available for creative use. A critical demand along this line is to predict the future product adoption for the merchants, which enables a wide range of applications such as targeted marketing. However, previous works only aimed at predicting if one user will adopt this product or not; the problem of adoption rate (or percentage of use) prediction for each user is still underexplored due to the complexity of user decision-making process. To that end, in this paper we present a comprehensive study for this product adoption rate prediction problem. Specifically, we first introduce a decision function to capture the change of users' product adoption rate, where various factors that may influence the decision can be generally leveraged. Then, we propose two models to solve this function, the Generalized Adoption Model (GAM) that assumes all users are influenced equally by these factors and the Personalized Adoption Model (PAM) that argues each factor contributes differently among people. Furthermore, we extend the PAM to a totally Bayesian model (BPAM) that can automatically learn all parameters. Finally, extensive experiments on two real-world datasets not only show the improvement of our proposed three models, but also give insights to track the effects of the various factors for product adoption decisions. IntroductionWith the help of information technology, the digital records of users' daily routines have provided an unprecedented opportunity to track the product adoption series of users. As a trend, leveraging these series for future adoption prediction has attracted increasing attention from both academy and industry [12,5]. Accurate prediction not only helps to understand the human decision process, but is also crucial to a wide range of business applications ranging from personalized recommendation [2], targeted marketing [4] to customer churn prediction [22].In the literature, many efforts have been devoted to the product adoption prediction problem, where the product could be a particular brand, a technology service or an opinion [10,23,8]. These works usually classified users in two categories, the adopters that have already consumed this product and the non-adopters that have not consumed it till now. Then classification methods were proposed to model the future adoption possi-

Recommending Profiles To Social Event Participation In Online Social Network Service Applications Using Deep Learning Techniques

“…RNN produces the hidden states of user profiles through the deterministic probability distribution using two properties, one is considered as distributed hidden state to store static social influence and social effects of the user and another is Non-linear dynamics to update dynamic social effects of the user. On incorporation of the linear constraints, back propagation algorithm on dynamic mutual influence can be modified easily between the weights of the social connection [10].…”