Research On Software Service Recommendation Method Based On Deep Learning

Since the beginning of the 21st century,with the rapid development of cloud computing,big data,artificial intelligence and other technologies,the Internet industry has continued to progress and mature.How to accurately recommend web services for target users in the massive Internet web services has become a hot issue in the field of service computing.In the current research on web service recommendation,the solutions for key aspects such as user feature extraction and web service semantic information extraction are not mature,and the impact of the dynamic characteristics of web services on the recommendation results is not well considered.Current research on web service recommendation unable to dynamically and effectively generate accurate and personalized recommendations for target users.Therefore,in response to the above problems,in the thesis,a software service recommendation method based on deep learning is proposed with the Programmable Web service platform as the data foundation and Web service as the research object.Specific research contents are as follows:Firstly,for the current research on web service recommendation,most of them are service recommendation based on Qo S,without considering the influence of user preferences on the accuracy of service recommendation,and the extraction of semantic information from web services is not yet mature.Therefore,the user's implicit feedback information is used to extract the user's interest preferences in this thesis;the relationship between API and Mashup is considered,the implicit relationship between Web services and Web services is explored,and the topic tags of Web services are supplemented by this thesis;the convolutional neural network recommendation model is migrated to the web service recommendation environment by this thesis,and its recommendation effect is verified by the Programmable Web dataset collected by crawler technology.Secondly,for the current research on web service recommendation,the dynamics of web service recommendation are not yet mature.The public web service recommendation data set does not take into account the timeliness of data.Convolutional neural networks have obvious advantages in local feature extraction,but they are slightly insufficient in understanding context information.Therefore,aiming at the problem of service dynamic perception,a service dynamic perception method is proposed.The BERT language representation model is used for feature vectorization;the recommendation model combining convolutional neural network and long short term memory network is adopted in this thesis to generate dynamic recommendation schemes for target users.The validity and accuracy of the recommended results are verified by the Programmable Web dataset collected by this thesis through web crawling technology.Finally,design a software service recommendation system based on deep learning.For the scalability of the system,different recommendation methods will produce different recommendation effects under different recommendation backgrounds and recommendation environments.The system can replace the recommendation model according to customer needs to meet the individual needs of users.Considering that the language training model and the deep neural network model may have corresponding problems in the accuracy of the recommendation results,multiple language training models are compared,and the recommended ideas of the above models are improved by this thesis.The experimental results show that the software service recommendation method based on deep learning proposed in this thesis can solve the problems existing in the current research on web service recommendation.The user's personal preferences are considered,the Mashup information is used to enhance the semantic features of the Web,the dynamics of the service are considered,and related deep learning models are proposed.The data set obtained is used for verification in this thesis,which proves that the method in this thesis comprehensively improves the effectiveness and accuracy of the recommendation.