Personalized Recommendation And Potential Value Prediction Based On Jin Gong Bao's Data

Government public resource trading websites all over the country will publish some information about bidding announcement,bidding candidates and other information.Many researches can be done bases on JinGongBao Data. This thesis mainly helps Companies to solve two difficult problems 1.Help companies find the bid that they are interested in 2.Help companies analyze the competitive of a bid. Therefore,the main work of this thesis as follows:Recommend different bid to different companies that they may be interested in which belongs to recommending domain.When a new bid published,using the recommendation algorithm based on frequency and the recommendation algorithm based on two part graph to predict the reference companies according to their history bidding data.When the length of seed is 10,the length of recommend list is l*5.The Recall of two algorithms can reach to 63.53%,62.53,the average length of two algorithms can reach to 27.53%,31.04.Help the company to analyze the competitive of a bid,which belongs to the regression problem.According to the actual situation,the appropriate objective function was selected,and the gradient descent optimization algorithm is used to solve the parameters.The top 10% of average rank accounted for 43.33%,the top 20% accounted for 57.66%,and the top 50% accounted for 88%.In general,the results of the forecast can reflect the competitiveness of most of the companies.