Research On The Recommendation Technology Of NQI Comprehensive Service Information System Based On Deep Learning

The National Quality Infrastructure(NQI)provides a strong guarantee for the healthy and orderly development of the national economy through measurement,standardization,inspection and certification,etc.However,the current regional development of NQI services in our country is uneven,and there are huge differences in service levels in different regions.The NQI integrated service information system is based on the bilateral resource integration model,integrating relevant NQI resources into the service system for users to choose and use,which greatly alleviates the problem of demand imbalance;but at the same time,the massive service data also leads to the problem of users' selection about service.To tackle this problem of the NQI system,relying on the NQI key research and development project "NQI integrated service information strategy and information system research and development",a research was conducted in this thesis about NQI service recommendation methods to improve the accurate retrieval performance of the NQI system.The thesis mainly carried out the following work:(1)Investigate and analyze the current status of domestic measurement services,standard services,inspection and testing services,and certification and accreditation services,summarize the existing problems at this stage,analyze the recommended requirements of the NQI integrated service information system,and present the recommended technical solutions for NQI services.(2)Propose a NQI service recommendation model based on deep collaborative filtering.The service recommendation method based on collaborative filtering only uses the interaction data between users and services with data sparseness,which limits the further improvement of the recommendation effect.This service recommendation model uses a heterogeneous information network to model the data heterogeneity with the advantage of flexibility.A NQI service heterogeneous information network is constructed using the interaction data between users and services,the type of service,the service organization and other attribute information,and the user and service information extracted from the NQI service heterogeneous information network;Through the deep neural network,further features are extracted from the user and service information,and the feature vectors obtained are sent to the fully connected network layer for NQI service recommendation.The thesis optimizes the parameters of the NQI service recommendation model to obtain the best parameters.Compared with the traditional Item Pop model,Item KNN model,matrix factorization model MF,matrix factorization model DMF based on deep neural network,etc.,the click-through rate HR and standardized discount cumulative gain NDCG of the deep collaborative filtering recommendation model proposed in this thesis increased by 6.36% and 3.97%,which shows the improvement of the recommendation effect.(3)Design a NQI service recommendation system.The in-depth collaborative filtering recommendation method proposed in this thesis is applied to the recommendation search of the NQI integrated service information system and tested.The test results show that the NQI service recommendation method designed in this paper meets requirements of the NQI integrated service information system.