A Composite User Preference Model:An Experimental Model From A Cognitive Perspective

Embarking on the Big Data Era,information utility has experienced exponential growth over the years.While the accessibility to information has greatly improved through increased interconnectivity to the World Wide Web and its citizens,the human mind is limited in capacity.Information overload is a serious concern.Information cannot be utilized fully,but is inevitably subject to compromise.Therefore,personalized service systems seek to alleviate this concern.The user preference,a key component of the personalized service system,aims to identify and extract an individual's likes,needs,aspirations,accurately through their Web-browsing behavior.Constructing a scientific and precise user preference model is the chief aim of field research.The current mainstream direct user preference modeling methods,are primarily based on registration information,satisfaction ratings,and user feedback.Indirect models make use of independents such as website visit time,save,bookmark,as well as the collection of hidden information.Rarely taken into account are the users psyche and the unpredictability of psychological states.The users cognitive state is left undiscovered and therefore their needs cannot be accurately determined.Previous studies on user preference and cognition have introduced Ebbinghaus'landmark forgetting curve from the field of memory research to update the user profile.Memory as the basis for process has witnessed the interest attenuation factor algorithm being devised to solve the problem of interest drift.However to update preferences by overemphasizing the effect of memory will fail to accurately reflect the temperamental nature of the human mind.User preference may be affected by other psychological factors.An over reliance on memory as an investigatory tool subjects the model to bias and corruption.Construction of the users should be subject to a host of variables beside memory.Instead,research scientists have taken the assumption that the degradation of memories will have substantial determined effects on preference without sufficient experimental data.There is a concern that without evidential cause for doing so,the strength of the assumption is weak.This paper seeks to address that concern by testing the relationship between memories and preference in real world data.The relationship between memories,affection and preference was clarified;borrowing from existing research.Inspired by the multi-store model,a composite user preference model appetite for real-time preference,short-term and long-term preference was built.Preliminary interviews were conducted on the browsing public to investigate the user's cognitive state in the website browsing process.Our experiment discovered that affection and memory are closely related;showing strong congruence in the first 7 days after exposure.Accordingly,the model establishes the short term preference threshold value of 7 days.The preference extraction method was realized.This model fully reflects the dynamic characteristics of user preferences,and provides a new perspective and approach for the main concerns of personalized services;data sparse and cold start.