Time-aware Service Value Based Service Recommendation Methods

With the flourishing of the new information technology,intelligent business services crossing network,domain and world have emerged in the modern service domain,which lead to the more complex service ecosystem.In such complex service ecosystem,how to quickly and effectively design service solutions for the specific requirement of users to meet the personalized needs of users and maximize user value has become a difficult problem.Service recommendation and service selection play an important role in the construction and optimization of service solutions.Service value is the ultimate goal of service recommendations.At present,most researchers pay attention to static value and rarely consider time-aware service value.To solve these issues,this thesis focuses on how to consider time-aware service value for service recommendations in the complex service ecosystem,so as to facilitate the rapid construction of service solutions.However,the diversification user records bring new challenges to the perception of time-aware service value and service recommendations.To solve the challenge,this thesis selects three familiar types of user records as representatives,and focuses on the research of the issue that sensing time-aware service value and recommending services from different user records.For the issue that the existing service value model cannot support the time-aware service value,a novel model named time-aware service value model has been proposed.This model proposes service value feature to describe user value preferences in fine-grained,based which user satisfactions will be calculated.Due to the time-aware user value preference,the user has different satisfactions with the same service at different time slots,which help complete the measurement of the time-aware service value.Compared with the existing service value model,such model is more practical,can avoid the harsh requirements for asking users to know all the service information,and improves the user experience.If given users’ service usage records,then a method defined as QoS-based timeaware service value sensing and service recommendation has been proposed.This method introduces service contract,which can more finely and comprehensively delineate users’ value preferences on numerical and non numerical attributes.Such method puts forward a user value preference sensing method by combining with statistical method.An evolution prediction algorithm for numerical attributes and non numerical attributes and a service recommendation method called QoS-based time-aware service recommendation are proposed.The analysis and prediction of time-aware user preferences can effectively alleviate the problem of unclear user requirements in traditional service recommendations,so as to improve the quality of service recommendations.Through a series of experiments based on real data,the effectiveness of QoS-based time-aware service value sensing and service recommendation method is proved.If given user behavior records,then a method named behaviors-based time-aware service value sensing and service recommendation is put forward.Such method proposes a personal service ecosystem model and the personal service ecosystem construction method based on user behavior records.This model intuitively and vividly describes the relationship between users and services,user behavior habits and other information.Meanwhile,this model effectively solves the singleness of the existing research while analyzing user behavior habits.A time-aware service value sensing method based on personal service ecosystem is put forward,and then this thesis analyzes the evolution of personal service ecosystem,which will help understand the time-aware service value.In addition,a method defined as user behaviors-based time-aware service recommendation method is offered,which alleviates the problem that the existing service recommendation methods heavily rely on other users,fully considers users’ personalized value preferences,and improves the quality of service recommendations and user experience.Through the collection of real data and a series of experiments,the effectiveness of the behaviors-based time-aware service value sensing and service recommendation method is proved.If given user review records,then a method named user reviews-based time-aware service value sensing and service recommendation is proposed.This method puts forward a novel service value feature extraction algorithm,which ensures that service value features have better practical significance and enhances the interpretability of service value features.The extracted valuable information is stored by knowledge graph.Thus,a domain-oriented user and service interaction knowledge graph is put forward,which naturally and intuitively shows the fine-grained relationships between users and services,and effectively alleviates the problem of data redundancy caused by the matrix storage structure.Based on the graph,adopting the graph structure search idea,five service recommendation algorithms are proposed.Relevant experiments based on real data have proved the effectiveness of the user reviews-based time-aware service value sensing and service recommendation method,which not only ensures the quality of service recommendation,but also greatly reduces the time complexity of recommendation methods.If the user requirement is complex,then a novel time-aware service value-based composite service recommendation method is proposed.For the global demand constraints of users,a demand decomposition method is given.Then,it builds a service solution that meets the personalized requirements of a user and maximizes the user value.Through the demand decomposition,it effectively solves the problem that users donot know much about candidate services and cannot express their requirements clearly.Through a series of experiments based on real data,the effectiveness of the time-aware service value-based composite service recommendation method is proved.On the premise of ensuring the quality of service solutions,the efficiency of the algorithm is greatly improved.In short,this thesis focuses on the service recommendation problem considering the time-aware service value in complex service ecosystem.It puts forward corresponding methods for different types of user records,which improve the recommendation accuracy and lay a foundation for building a service solution that meets the personalized requirements of a user and maximizes the user value.