Research On Multi-dimensional Quality-driven Service Recommendation And Privacy Protection

With the rapid development of Internet technology and the advent of the era of big data,the scale of Internet users is expanding rapidly and the amount of information is expanding rapidly.On the one hand,massive information meets people's needs in all aspects of life.On the other hand,a large number of mixed information makes users unable to obtain the required information timely and accurately.Therefore,while people are enjoying the convenience of life brought by the era of big data,they are also facing the serious problem of "information overload".In this case,the emergence of recommender system has become one of the effective ways to solve the problem of information overload,which is essentially an information filtering system.Through the analysis of user's historical behavior,it can mine user's behavior preference,and then help users filter out the useless information in the mass of information,and recommend the information in line with user's preferences.Because of the independence and easy explanation of the domain,collaborative filtering has become one of the effective methods to realize service recommendation.However,there are still some shortcomings in traditional collaborative filtering algorithms.For example,the efficiency of recommendation methods is not high,it is difficult to meet the high response needs of users and effectively protect the user's privacy information,which is easy to cause users' concern about the disclosure of privacy information.Nowadays,Locality Sensitive Hashing(LSH)technology is applied in the field of service recommendation to protect users' QoS privacy data and make up for the shortcomings of traditional collaborative filtering recommendation algorithm.However,traditional LSH-based recommendation algorithms still have some problems.Specifically,(1)The kind of methods has some limitations in the case of sparse QoS data,so it cannot solve the problem of sparsity and cold start.(2)At present,the research of LSH-based recommendation methods is limited to single dimensional quality data(such as response time or throughput),but does not consider the multi-dimensional quality data(such as response time and throughput)and the possible correlation between different quality indicators.Therefore,it is not conducive to improve the applicability and accuracy of the service recommendation system,but also leads to the unreasonable recommendation results.In view of the above shortcomings and challenges,this paper proposes a multi-dimensional quality driven privacy aware service recommendation method MD-LSH and a multi-dimensional quality driven relevance aware service evaluation method MD-TOPSIS.The specific research contents are summarized as follows.(1)MD-LSH,a service recommendation method proposed in this paper,mainly improves traditional LSH-based recommendation methods to make it suitable for multi-dimensional quality driven service recommendation scenarios.Specifically,MD-LSH first analyzes the user's multi-dimensional historical QoS data to construct the user service multi-dimensional quality matrix;then uses LSH to construct low privacy or no privacy user index matrix offline,and uses the user index matrix to find similar users;finally,it predicts the service QoS and generates the candidate service recommendation list.Experiments show that the proposed MD-LSH method can improve the accuracy,applicability and efficiency of the service recommendation system on the premise of protecting user privacy,and overcome the part of the cold start problem caused by data sparsity,thus making the system more robust.(2)MD-TOPSIS,a service evaluation method proposed in this paper,introduces Mahalanobis distance to eliminate the correlation between multi-dimensional quality data,and combines with TOPSIS(technology for order preference by similarity to ideal solution)to comprehensively evaluate and sort the candidate service list generated in MD-LSH.Specifically,MD-TOPSIS constructs the matrix to be evaluated according to the candidate service list to determine the positive ideal solution and negative ideal solution;then,it uses Mahalanobis distance to calculate the distance between the service to be evaluated and the positive ideal solution and negative ideal solution;finally,it calculates the utility value of each candidate service,and sorts the services in the recommendation list according to it,so as to select the best service to recommend to users.Through case study,the feasibility of MD-TOPSIS method is proved.