Research On Recommendation Algorithm Of Science And Technology Resources Based On Deep Learning

The science and technology service industry is a part of all service industries in today's society.There is an urgent need to use the technology service industry to promote the rapid development of the regional economy.However,research on science and technology resource services is still relatively small,and the amount of science and technology resources is huge but chaotic.Therefore,integrating regional high-quality technical service resources,building a comprehensive technical service cloud platform for Beibu Gulf urban agglomerations,and applying recommendation algorithms to technical resource recommendation is of great significance for the development of Beibu Gulf urban agglomerations and the research of technology practitioners.Based on the above research background and the hot position of deep learning in the recommendation industry,this article studies the technology resource recommendation algorithm based on deep learning.The main research contents of this paper are as follows:(1)Aiming at the problem of large amount of scientific and technological resources and sparse data,the DeepFM model that can capture high-order linear features and low-order linear features is applied to the scientific and technological resource recommendation algorithm,and the fusion of FM model and DNN model is used.,Using the FM model to deal with the problem of low-level feature information and data sparseness in scientific and technological resources,and using the DNN model to deal with the problem of high-level feature information in scientific and technological resources,effectively solving the problem of data sparseness and the high amount of data in scientific and technological resources.This has been proved through experiments.Compared with the separate FM model and DNN model,the DeepFM model has a better overall effect in the scientific and technological resource recommendation algorithm.(2)Propose a VAE-GAN technical resource recommendation model.A recommendation model is studied,which generates a confrontation network while adding help information to the previous variant autoencoder.The original variant autoencoder can make use of the user's ever-increasing information or scientific and technological resources after innovation.Other information is used to calculate the parameters of the distribution,replace the imperfect Gaussian distribution,expand the latent variable space,and more accurately show the user's preferences.In the process of training the model,the discriminant model is integrated to improve the measurement standard of the variational autoencoder,in order to better improve the effect of the scientific and technological resource recommendation system.Finally,experiments have shown that the Recall and NDGG indicators of the model have achieved good results.(3)In order to verify the effectiveness of the VAE-GAN model,a scientific and technological resource recommendation system based on the VAE-GAN model is established.The system can recommend relevant scientific and technological resources for users,and adds a project discussion module to participate in the Beibu Gulf project.This person provides conditions for communication and improves the practicability of the system.