Research And Implementation Of Preference Query On CP-nets And TCP

With the rapid development of Internet technology,the phenomenon of information overload has become the biggest obstacle for people to obtain the information they need.Preference queries are important types of preference inference learning.It can provide users with personalized services according to their needs,and return results corresponding to their needs to users more conveniently and efficiently.It has important academic value and broad application prospects in recommender systems,multi-criteria decision-making and computer science theory.The traditional research on preference query mainly focuses on the preference of a single object represented by relational tuples,and most of them are oriented to static relational databases,while the method of preference query is extended to sequences representing users' multi-attribute preference requirements and sequence data flow dynamics.The scene remains a challenge.Therefore,based on the existing research on preference query,this paper proposes a learning method for generating consistent ranking of CP-nets(Coditional Preference Networks)based on preference priority in a static environment and a data flow sequence extracted based on temporal conditions in dynamic scenarios.Preference query learning method,the main research work of the paper is as follows:(1)A learning method for generating consistent ranking of CP-nets based on preference priority is proposed.In order to solve the complex needs of users with multiple attributes,this paper starts from the basic learning method of conditional preference network CP-nets,a qualitative preference inference tool,and obtains the complete preference relationship between user attributes by consistent sorting of CP-nets.And use this method to design a consensus ranking generation algorithm,which is used to quantify the preference of each result attribute of the user,and obtain the preference priority of the result attribute.These values can accurately reflect the user's preference,so as to obtain the arbitrary structure of CP-nets.Consistent ordering.At the same time,this paper also shows the process of optimizing the CP-nets consistent ordering update.Finally,it is verified by experiments that the consistent sorting algorithm can not only obtain the complete sequence of user's multi-attribute requirements,but also have a faster inference speed.(2)This paper proposes a query learning method for data stream sequence preference based on temporal condition extraction.Facing the data flow in real-time applications,this paper combines the conditional preference theory with the time series data flow scenario.Firstly,a time condition preference query statement Stream Seq(Stream Sequence)is proposed,which is used to process the time condition preference on the data stream and perform time condition preference reasoning on the sequence extracted from the data stream.Then,an extraction sequence algorithm for processing objects in the data stream according to time conditions and a dominant comparison algorithm for obtaining a dominant sequence according to the extraction sequence are proposed.Finally,the method is experimentally verified by synthetic datasets and real datasets.The experimental results show that,compared with other methods of preference query,the method in this paper requires a shorter running time and is more efficient in obtaining results.In summary,the CP-nets consensus ranking generation algorithm effectively solves the problem of how to efficiently obtain the complete sequence of users' multi-attribute preference requirements.In addition,in addition to the static rules used in CP-nets,the use of qualitative conditional preferences to normalize and reason about time-indexed dynamic data flow sequences can express the impact of user preferences in past moments on preferences in current and future moments.Finally,the work of this paper is summarized and the direction of future work is given.