Research On Hierarchical Web Service Classification Method Based On Deep Learning

Nowadays,many web service providers publish their business services in the cloud,which leads to a large number of Web services with similar functions growing rapidly,making web service discovery more challenging.In the existing service registries,web services are usually described by web services description language or simple natural language text.How to accurately and efficiently retrieve web services that meet the needs of users has become a hot issue.Classifying web services with similar functions is an effective way to promote web service discovery.Most of the early service classification methods are based on the traditional machine learning model,which need difficult Feature Engineering,and the ability of feature representation is weak.Moreover,the classification method ignores the hierarchical structure of web service category,and is only suitable for single-layer category classification.Even though the current service classification methods based on deep learning model avoid the thorny Feature Engineering and have strong feature representation ability,they still have not solved the hierarchical web service classification problem.Based on the increasing number of Web services,the scale of recommendation results obtained by simple one-time classification is still large,which can not meet the needs of service requesters to quickly locate the required services through one-time classification.Based on this problem,this paper proposes a hierarchical classification method to quickly narrow the search range of service requesters,which greatly reduces the service discovery time.For the first time,this method applies an ordered long-term and short-term memory neural network(ONLSTM),which can extract the syntactic structure of text,to the field of service classification.Combined with hierarchical fine-tuning and parent class embedding technology,it fully excavates the constraint relationship between class levels.In order to prove the effectiveness of this method,this paper constructs a data set with 10184 web services and two-tier categories,and compares the classification accuracy of 13 classification models.Experimental results show that the proposed model has the best classification effect compared with other classification models.