Research And Implementation Of QoS Aware Recommendation Algorithm In Heterogeneous Environment

With the increasing demand for information,internet platforms have produced a large number of web services with diverse functions.Faced with a massive amount of service information,users often need to spend a lot of time searching for services that meet their own needs.In order to improve the timeliness and accuracy of web service invocation,service recommendation algorithms have emerged.However,in practical application scenarios,services are distributed in different edge servers,resulting in heterogeneous environments with different distribution of Quality of Service(Qo S)for users.On the one hand,the aforementioned heterogeneous environments cause changes in the interaction data between users and services over time and network dynamics,increasing the complexity of service recommendations.On the other hand,it leads to scattered Qo S data generated by users calling services and poses a risk of privacy leakage.Therefore,in order to recommend efficient and accurate web services to users in heterogeneous environments,the paper explores the above issues and proposes a single platform service recommendation algorithm based on time perception and a cross platform data feature aware service recommendation algorithm based on privacy protection.In addition,to verify the scalability of the proposed algorithm,the paper further extends the algorithm to practical application scenarios to solve traffic congestion problems in daily life.The specific work is as follows:(1)A time aware service recommendation algorithm is proposed to address the dynamic variation of Qo S values when users call web services on a single platform.This algorithm introduces a time dimension to break through the limitations of the uniqueness and stationarity of user and service interaction data.It is mainly divided into two stages: the first stage predicts the missing values in the historical time slice,and uses the memory based collaborative filtering recommendation algorithm to predict the missing Qo S values in the historical time slice.The second stage analysis found a strong correlation between user Qo S values over continuous time gaps,and predicted missing Qo S values in the current time slot by referencing Qo S values in historical time slots.Compared with other algorithms,this algorithm not only ensures accuracy and timeliness,but also meets the needs of users to dynamically call services.(2)A privacy protection based cross platform data feature aware service recommendation algorithm is proposed to address the sparse user service interaction data and user privacy security issues in cross platform data.The model is divided into two stages: privacy protection and prediction of missing Qo S values.In the first stage,Security Policies and Procedures(SPP)are used to encrypt user data,achieving the goal of protecting user privacy.In the second stage,the Latent Factor(LF)model is used to densely extract sparse data,and density based clustering is performed on the data to predict missing values through matrix decomposition.Use the same prediction algorithm for other platforms with user information,and then perform weighted summation to obtain the final prediction result.Compared with other algorithms,this algorithm not only protects user privacy,but also to some extent alleviates the cold start problem caused by sparse data,improves the accuracy of service recommendation,and distributes the predicted results to the edge,reducing service response time.(3)Integrating and promoting the proposed web service recommendation algorithm into smart transportation application scenarios,a smart transportation road prediction and recommendation system has been designed and implemented.The system considers the impact of time perception and cross platform data feature perception on path prediction in heterogeneous environments,providing users with appropriate road services to avoid traffic congestion to a certain extent,and further verifying that the algorithm proposed in the paper has certain scene scalability.