Research On Prediction Method Of User Influence And User Dynamic Behavior Of Online Review Website

With the development and popularization of Internet and mobile terminal technology and the maturity of online interactive platform technology,all kinds of online review website services have been extended to all aspects of people's life,covering the daily life of clothing,food,housing,and transportation.In particular,catering online review websites have become a platform with the highest user penetration rate and utilization rate.The catering industry is gradually developing towards e-commerce,and the research on e-commerce in the catering industry has become a hot spot of frontier interdisciplinary research.It has become a popular trend for users to search online review websites before eating,collect and compare information,and obtain useful information to make dining decisions.However,with the e-commerce of the catering industry and the speed of development accelerating year by year,online review websites not only bring convenience to users but also bring them the trouble of information overload.For users,how to effectively screen information from massive data and obtain useful information is the primary demand of users and also the main problem they are facing at present.For businesses,how to make their restaurants stand out among many businesses and make their restaurants recognized by more users is the main dilemma facing most businesses at present.For the platform,maintaining the popularity of the platform and promoting the sustainable and healthy development of the platform are the most fundamental needs.In order to solve the above problems,more and more scholars use data mining,machine learning and other methods to solve the problems encountered in the e-commerce of the catering industry.For the above food industry problems in the process of development of electronic commerce based on Yelp online review sites as the research object,in order to reduce user dining decision cost,improve the platform to users dining behavior dynamic prediction ability,and enhance platform can recommend systemic main goal,for online review sites users influence and eating behavior analysis.The traditional collaborative filtering recommendation model is improved.First of all,users with high influence can be calculated through the measurement of user influence,which can help the platform formulate reasonable incentive policies for such users to keep them active and attract more users to jointly maintain the popularity of the platform.At the same time,it can also reduce the time cost of information screening for ordinary users.Then,the dynamic change of users' dining behavior is analyzed to explore the geographical location and taste change trend of users and predict their dining behavior in the future,providing a basis for improving the performance of the platform recommendation system.Finally,the platform,merchants and users can achieve a win-win situation,which is conducive to promoting the e-commerce development of the catering industry and promoting the e-commerce model of the Yelp website to become more mature.The main research contents and achievements of this dissertation include:(1)This dissertation improves the existing measurement model of user influence.Based on the existing subjective weighting method with objective weighting method to determine the user influence measure of the weight calculation method,this study firstly by combining Yelp users attribute structure influence measure platform,puts forward the comprehensive weight assignment method to determine the weight of each measure,and then use the way of weighted summation,the influence to the user.Finally,the comparative analysis of the calculation results proves that the proposed method is better than the subjective weighting method and the objective weighting method alone,and is more helpful for the platform to measure user influence and identify high influential users.(2)This dissertation mines and analyzes the user's dining behavior information,studies the changing trend of the user's dining behavior,constructs the network of the user's dining behavior,and introduces information entropy to measure it.The dining behavior of users is studied and analyzed from geographic location information and dining taste information of users.Users dining location dynamic network is constructed and dynamic network user meal tastes,by information entropy measurement analysis found that the user's dining location and dining have boundedness and unreal the taste,the user's dining behavior predictability,this recommendation algorithm improvement provides the foundation for the next step.(3)This dissertation improves the traditional collaborative filtering algorithm and proposes a collaborative filtering recommendation model integrating user dining behavior information.Firstly,based on the information analysis of users' dining behavior,the dynamic network of users' dining behavior was characterized by vector representation,and the inertia coefficient of users' dining preference was defined.To solve the problem that traditional collaborative filtering algorithms only use user ratings for recommendation and data sparseness and cold start,this dissertation integrates the inertia coefficient of user dining preference and user ratings information into the traditional collaborative filtering algorithm to obtain an improved collaborative filtering recommendation model integrating user dining behavior information.Compared with other commonly used recommendation algorithms,the recommendation algorithm in this dissertation has improved significantly.