Design And Implementation Of Recommender System For Expert Based On User Evaluation

In China,the technological innovation and reform problems of enterprises are obviously serious.Middle and small enterprises,as one crucial part of industry-university-research cooperation,need adequate supports of funds,technologies and talents from the government and universities.Industry-university-research cooperation has got many achievements in China,but it also has many unaddressed issues,such as low rate of technical conversion,power shortage,low system efficiency and so on.Under this situation,enterprisers has troubles in finding suitable experts for technical consultant when facing problems of technological innovations and rechanges.In order to solve this problem,this thesis design and implementa a expert recommend system for enterprise based on data mining,research and analysis the similarity between the technical requirement text of enterprise and the dissertation text of university expert,build the model for the user evaluation system and the model of recommender hybrid system,to implementa the expert recommender system based on user evaluation.The major research of the thesis is as follow blow:(1)Getting the expert data of university by the web crawler form CNKI,through the word segmentation,filter the stop word,keywords extraction to the abstract text collection of the expert.Then we use the space vactor methon to express keywords and their weight points,building the space vector model for professor.(2)In view of the enterprise technology requirement text characteristics,after segmentating the word and filtering the stop word to the requirement text,combine the keywords extraction methods which are based on word semantic web and word statistical information.Then we build the space vactor model for the enterprise technology requirement text,and build the Classification model to classify the requirement text.At last we calculate the similarity between the text imformation and the expert dissertation text,give the expert recommendation list according to the similarity.(3)Buliding the user evaluation system.According to target user's satisfaction rating of the initial expert recommendation list,and combining the historical rating data,the system calculates the target user's prediction satisfaction rating score for each expert.At last the system adjust the initial expert recommendation list and give the finally expert recommendation list according to the prediction score of satisfaction.(4)In this thesis,a new expert recommender system model based on user evaluation is proposed.This system adopts the hybrid recommender model,at first the system calculates the cosine similarity between the enterprise technology requirement text and the expert dissertation text,and give initial recommendation result based on content-based recommendation algorithm.The commendation result takes as the input of the user-ratings recommendation algorithm,combining users satisfaction ratings,produce the finally expert recommendation list.Finally,based on those study results above,the expert recommender system is achieved.And this thesis use lots of actual data to test the system.The result shows that the recommender system is reasonable and exact.And it also shows that the expert recommender system model is scientific and useful.