Search Engine Based On RBM

As the rise of Internet and mobile Internet,the interactive degree between users and Internet is becoming deeper and deeper.We can gain an insight into the users’ experience,business marketing,personal tastes,and human behavior which we are familiar with by a cetain amount of data of users’feedback.Thus we can provide a more targeted services.However,with the growing wealth of information,what users widely focus on is how can they get the information more accurately and rapidly through the search engine,and the researching of page rank algorithm through it which is also one of the hotspots.The traditional page rank algorithm is based on the content.In order to improve the page rank and attract more people,some one would like to add inrelated content or use the other way to cheat on them.So how to use the data of users’feedback to improve the quality of page rank efficiently and to prevent from the act of cheating is what we focus on this research.The research work is as follows:(1) Aiming at the cheating problem in the Page rank algorithm, research put forward an page rank algorithm based on Restricted Boltzmann Machines(RBM) is proposed.The algorithm conbines the pattern recognition of RBM, with the data of users’feedback,it could adjust the correlation coefficience between words and pages and improve the results of page rank. The results of experiment indicated that not only could the algorithm reverse the results of the high related retrieves requests back for a new one reasonably based on the feedback data,so it could prevent the rank cheating effectively,but also it is capable of calculated about the inputs that they have never done before.(2)When users retrieves information,they will pay more attention to that whether the results meet their needs or not,and put forward personalized search engine based on RBM.The algorithm will train the data that combines with user’s personalized information and adjust the correlation coefficient.When the user retrieves information,the page rank would be sroted by adjusted correlation coefficient,and meet the users’individual needs.(3)The process of page rank and training has seen the time consuming problem, put forward an search engine based on concurrency is proposed.According to introduce the compute unified device architecture(CUDA),each step of page rank and training could achieve inteanal parallel computing,and the technology of stream in cuda will make the page rank and training execute concurrently.Using the simulation technology for search engine with the process.The result shows that CUDA could accelerate the process,increase the inputs and outputs per unit time.Accelerating decreases waiting time for user, timely update reflect the feedback of user quickly,and the results are more accurate.In domestic,under the condition of shortages in researching on the data of user feedback,this paper combines RBM and search engine that based on user feedback, establish the search engine based on RBM model.and achieving the page rank algorithm.The results of research will make a foreward to improve the page rank algorithm with the data of feedback and to offer a targeted sevices with a important theroical reference value and use value.