Probabilistic graph model and neural network perspective of click models for web search

Click behavior is a typical user behavior in the web search. How to capture and model users’ click behavior has always been a common research topic. However, there are few review studies on this topic. In this paper, we present a survey to comprehensively analyze click models of web search via types of models. Based on differences in research hypotheses and modeling methods, click models are generally divided into probability graph-based click models, neural network-based click models, and hybrid click models. Firstly, we give a discussion of click models in the extant literature, within which basic assumptions and their extensions, advantages and disadvantages for click models are presented. We also compare and analyze the characteristics and application scenarios of different types of models. Secondly, we choose eight representative click models and conduct comparative experiments on two real-world session datasets to compare their performance. Finally, we identify current research trends, main challenges and potential future directions of click models worthy of further explorations.