The Design And Implementation Of Real-time Crowdsourcing Oriented Mobile Vertical Application

As a new type of social production mode,crowdsourcing has greatly integrated social resource and created tremendous commercial value.With the development of internet technology,crowdsourcing has developed into a social behavior,which was used to be a purely business practice.Confront the rise of mobile internet,there is not only an opportunity but also a great challenge for crowdsourcing.Information in existing crowdsourcing applications is usually represented in category directly,which led to the information overload problem,it’s difficult for the application to provide a quick response for users’ real-time collaborative task in mobile scenarios.Based on practical group buying demand,this thesis solve this problem using recommendation system.The primary goal of the application is to design a recommended engine architecture which could not only support complex algorithm,but also respond requests in real-time.Based on distributed event system Apache Kafka,distributed parallel computing framework Apache Spark,distributed memory database Redis and distributed NoSQL database HBase,a recommend engine is designed and implemented in this thesis,which contains three parts:online,near-line and offline.The offline part executes batch calculation,the near-line part executes increment calculation and the online part generate result instantly,it’s been proved to be practical in the thesis.Most existing activity recommend algorithms are making use of traditional algorithm,which ignore the characteristics of activity itself.The complexity of mobile scenarios provides a lot of useful information,such as location and social relationship,which constitutes an event-based social network(EBSN).As a new hotspot,EBSN is receiving more and more attention.Based on the research of influence between user’s attendance of activity with semantic attribute,location attribute and social attribute,the thesis proposes a latent-neighbor-factor-based singular matrix decomposition algorithm.Compare with traditional prediction model,this model can effectively alleviate the cold start problem and predict user’s attendance more accurately.Based on the above recommend engine and recommend algorithm,the thesis designs and implements a mobile application which could provide group buying service.Considering user’s interests and real time context,intelligent recommendation service is provided in the application,which could effectively alleviate the information overload problem.