Research On Technology Service Resource Recommendation Method Based On Enterprise Portrait

As an important part of China's modern service industry,science and technology service industry is an important support to promote scientific and technological innovation,transformation of scientific and technological achievements.It is of great significance for improvement and optimization of industrial structure,and improvement of economic quality and economic efficiency to integrate the advantageous science and technology service resources of relevant regions and build a comprehensive science and technology service cloud platform at the same time.The information of science and technology service resources is growing infinitely,but the energy of enterprises to find more suitable resources is limited.Therefore,how to use an efficient and accurate recommendation algorithm to recommend technology resources for enterprises has become an important issue to improve resource utilization and service efficiency.This thesis focuses on how to use the corporate portrait to recommend the necessary science and technology service resources to relevant science and technology enterprises.Firstly,based on the generation of dynamic enterprise portrait of multi-source heterogeneous data,through the pre-processing of multi-source heterogeneous data,the enterprise information is described and integrated;This thesis proposes a model of multi-dimensional representation of enterprise,which uses natural language processing to realize feature analysis and label extraction,and depicts the enterprise portrait from multi-dimensional;A dynamic updating method of enterprise portrait is designed,which extracts the keywords and weights of the new change information of the enterprise,and judges the extracted features.If it is new,it will be used as the new feature item of the enterprise portrait and add the corresponding weight.If it is old,it will update the enterprise portrait according to the updating formula,and finally generate the dynamic description label with weight.Secondly,aiming at the enterprise portraits,enterprise feature tags and their weights generated from multi-source heterogeneous data,this thesis proposes an inverted information matching similarity calculation method by using the portraits data to recommend technology service resources to enterprises,which reduces the complexity of calculation time and introduces the influence of weight,so that the matching degree between enterprises and technology service resources is higher;Combined with the current mainstream matrix factorization collaborative filtering algorithm based on hidden semantic model,the enterprise portrait tag recommendation algorithm and matrix factorization collaborative filtering algorithm based on hidden semantic model are integrated in different ways,and the weight factor is determined by MAE evaluation method.Finally,the recommendation algorithm based on enterprise portrait tag(EP),matrix factorization collaborative filtering based on hidden semantic model(CF),hybrid recommendation algorithm based on union of the above two algorithms(m-EPCF)and collaborative filtering algorithm based on hidden semantic model are proposed ? Hybrid recommendation algorithm based on weighting factor(?-EPCF).The results show that based on the ? Hybrid recommendation algorithm based on weighting factor(?-EPCF)is superior to the other three algorithms in terms of comprehensive accuracy,recall and F-measure.Therefore,this thesis proposes a method based on ? Hybrid recommendation algorithm based on weighting factor(?-EPCF)can not only effectively use enterprise portrait and technology service resource tags,but also reduce the impact caused by the lack of scoring information and sparse matrix.It verifies that this method can discover enterprise users' interest in technology service resources through enterprise portrait,and the recommendation algorithm improved according to this method can effectively improve the accuracy of recommendation.