Service Prediction And Recommendation Methods Based On Multi-source Heterogenous Personal Service Data

The ubiquity of cloud computing and the localization,mobility and socialization of Internet services contribute to the great prosperity of available services on the Internet.The "person" who use these services can be divided into "customers" in the service supply and demand relationship and "service providers" in the crowdsourcing service-tasks.The core issue to meet customers' demands is how to plan effective service schemes according to their personalized demands and seamlessly implement them with mobile applications and cloud services.The core issue to meet service providers' demands is how to recommend the appropriate crowdsourcing service-task to them.In order to meet the personalized and differentiated demands of customers and service providers,opportunities and challenges are presented for the current mediation-based service prediction and recommendation methods:(1)Mediation-based service prediction and recommendation methods encounter bottlenecks in deep personalization.(2)The media-centric data organization model is difficult to support the service prediction and recommendation based on the correlation of personal service data due to the dispersion of multi-source heterogeneous data.(3)Social preferences that service providers participating in crowdsourcing service-tasks are not considered,which lead to bottlenecks in the accuracy of recommending crowdsourcing service-tasks.To solve the above problems,this paper studies the methods of service prediction,service recommendation and crowdsourcing service-task recommendation based on multi-source heterogeneous personal service data generated by person in the process of using these service.The specific research work includes the following aspects:(1)Service prediction and recommendation-oriented integrated modeling personal behavior and data.Due to the current decentralized storage of personal-service data,correlated personal-service data is isolated,which makes it impossible to trace the correlation of using services based on the correlation of personal-service data.Therefore,the correlation between different personal-service data needs to be recovered firstly.A time-aware and multi-source heterogeneous personal behavior and service data integrated model is researched,which can aggregate the personal-service data scattered among different services by one person,restore the original correlation between personal-service data and break the isolation of them.Moreover,the model can follow and represent how personal-service data changes over time.The purpose of this model is to enable service prediction and recommendation methods to trace the correlation of services by using the correlation between personal-service data under the "user-centered data management mode” and then make the personalized service prediction and recommendation schemes.(2)Service prediction method based on personal-service data transition patternsand service behaviors.Predicting services that maybe used in the future for the "customer" in the service supply and demand relationship.The traditional service prediction methods only focus on the single service behavior or do not fully consider the situation that the variation of personal-service data can trigger the execution of related services.It is necessary to consider the impact of the variation of personal-service data on the customer's subsequent usage of services.A service prediction method based on personal-service data transition pattern and service behavior is researched,which considers the patterns of personal-service data transition and service behavior.When a user produces new personal-service data,the service prediction method matches the most likely services to use based on the transition patterns of personal-service data.In order to evaluate the effectiveness of this method,the historical personal-service data generated by real-world customers is collected.The proposed method and the service prediction method that do not take sufficient account of the data variation are running on the data set.The experimental results show that the proposed method can predict the service more accurately than the comparison methods.(3)Service recommendation method based on service behavior and personal-correlated data.Recommending various services for the "customer" in the service supply and demand relationship.The methods that can be used for service recommendation in related work do not consider sufficiently the correlation between different personal-service data.A deep recurrent neural network model based on personal service behavior and personal correlated data(these data is linked to each other)is researched,which breaks users' habit of using services,recommends a variety of services for users.In order to evaluate the effectiveness of this method,historical personal-correlated data generated by real-world users using the service is collected,and the data set is divided into training set and testing set.In the training set,personal-correlated data is used to train the neural network model.The trained neural network model and the methods that do not take into account the correlation between different personal-correlated data are used to recommend services in testing set,and the accuracy is compared.The experimental results show that the recommendation model based on personal service behavior and personal-correlated data is more accurate than the comparison methods.(4)Crowdsourcing service-task recommendation method based on personal social preferences.Recommending the most appropriate crowdsourcing service-tasks for the“person” involved in crowdsourcing of the group intelligence service,where the person is regarded as “service provider” of the crowdsourcing service-task.Traditional crowdsourcing service-task recommendation methods only focus on personal preferences that individual completes tasks alone.However,in the scenario of group intelligence crowdsourcing service development,the social preferences of individuals are not sufficiently considered in most related work.A crowdsourcing service-taskrecommendation method based on personal social preferences is researched to recommend the most appropriate crowdsourcing service-task to the service provider,which combines the personal social preferences.Research on the method of modeling personal social preferences,and apply the personal social preference model to the crowdsourcing service-task recommendation method to improve the accuracy.In order to evaluate the effectiveness of the proposed method,historical data generated by service providers participated in crowdsourcing service-tasks is collected.The crowdsourcing service-task recommendation method and methods that do not consider personal social preferences are running on the data set to verify the accuracy of the recommended crowdsourcing service-task.Experimental results show that the crowdsourcing service-task recommendation method based on personal social preferences is more accurately than the comparison methods.(5)Service prediction and recommendation prototype system.According to the personal behavior and service data integrated model,service prediction,service recommendation and crowdsourcing service-task recommendation methods proposed above,the service prediction and recommendation prototype system are designed and developed by utilizing the real-world personal-service data in the scenario of open source software development services application.