Mingdong Tang

An Effective Web Service Recommendation Method Based on Personalized Collaborative Filtering

Collaborative filtering is one of widely used Web service recommendation techniques. There have been several methods of Web service selection and recommendation based on collaborative filtering, but seldom have they considered personalized influence of users and services. In this paper, we present an effective personalized collaborative filtering method for Web service recommendation. A key component of Web service recommendation techniques is computation of similarity measurement of Web services. Different from the Pearson Correlation Coefficient (PCC) similarity measurement, we take into account the personalized influence of services when computing similarity measurement between users and personalized influence of services. Based on the similarity measurement model of Web services, we develop an effective Personalized Hybrid Collaborative Filtering (PHCF) technique by integrating personalized user-based algorithm and personalized item-based algorithm. We conduct series of experiments based on real Web service QoS dataset WSRec [11] which contains more than 1.5 millions test results of 150 service users in different countries on 100 publicly available Web services located all over the world. Experimental results show that the method improves accuracy of recommendation of Web services significantly.

Location-Aware Collaborative Filtering for QoS-Based Service Recommendation

Collaborative filtering is one of widely used Web service recommendation techniques. In QoS-based Web service recommendation, predicting missing QoS values of services is often required. There have been several methods of Web service recommendation based on collaborative filtering, but seldom have they considered locations of both users and services in predicting QoS values of Web services. Actually, locations of users or services do have remarkable impacts on values of QoS factors, such as response time, throughput, and reliability. In this paper, we propose a method of location-aware collaborative filtering to recommend Web services to users by incorporating locations of both users and services. Different from existing user-based collaborative filtering for finding similar users for a target user, instead of searching entire set of users, we concentrate on users physically near to the target user. Similarly, we also modify existing service similarity measurement of collaborative filtering by employing service location information. After finding similar users and services, we use the similarity measurement to predict missing QoS values based on a hybrid collaborative filtering technique. Web service candidates with the top QoS values are recommended to users. To validate our method, we conduct series of large-scale experiments based on a real-world Web service QoS dataset. Experimental results show that the location-aware method improves performance of recommendation significantly.

Web Service Selection for Resolving Conflicting Service Requests

Web service selection based on quality of service (QoS) has been a research focus in an environment where many similar web services exist. Current methods of service selection usually focus on a single service request at a time and the selection of a service with the best QoS at the users own discretion. The selection does not consider multiple requests for the same functional web services. Usually, there are multiple service requests for the same functional web service in practice. In such situations, conflicts occur when too many requesters select the same best web service. This paper aims at solving these conflicts and developing a global optimal service selection method for multiple related service requesters, thereby optimizing service resources and improving performance of the system. It uses Euclidean distance with weights to measure degree of matching of services based on QoS. A 0-1 integral programming model for maximizing the sum of matching degree is created and consequently, a global optimal service selection algorithm is developed. The model, together with a universal and feasible optimal service selection algorithm, is implemented for global optimal service selection for multiple requesters (GOSSMR). Furthermore, to enhance its efficiency, Skyline GOSSMR is proposed. Time complexity of the algorithms is analyzed. We evaluate performance of the algorithms and the system through simulations. The simulation results demonstrate that they are more effective than existing ones.

Location-Aware and Personalized Collaborative Filtering for Web Service Recommendation

Collaborative Filtering (CF) is widely employed for making Web service recommendation. CF-based Web service recommendation aims to predict missing QoS (Quality-of-Service) values of Web services. Although several CF-based Web service QoS prediction methods have been proposed in recent years, the performance still needs significant improvement. First, existing QoS prediction methods seldom consider personalized influence of users and services when measuring the similarity between users and between services. Second, Web service QoS factors, such as response time and throughput, usually depends on the locations of Web services and users. However, existing Web service QoS prediction methods seldom took this observation into consideration. In this paper, we propose a location-aware personalized CF method for Web service recommendation. The proposed method leverages both locations of users and Web services when selecting similar neighbors for the target user or service. The method also includes an enhanced similarity measurement for users and Web services, by taking into account the personalized influence of them. To evaluate the performance of our proposed method, we conduct a set of comprehensive experiments using a real-world Web service dataset. The experimental results indicate that our approach improves the QoS prediction accuracy and computational efficiency significantly, compared to previous CF-based methods.

Elastic Personalized Nonfunctional Attribute Preference and Trade-off Based Service Selection

For service users to get the best service that meet their requirements, they prefer to personalize their nonfunctional attributes, such as reliability and price. However, the personalization makes it challenging for service providers to completely meet users’ preferences, because they have to deal with conflicting nonfunctional attributes when selecting services for users. With this in mind, users may sometimes want to explicitly specify their trade-offs among nonfunctional attributes to make their preferences known to service providers. In this article, we present a novel service selection method based on fuzzy logic that considers users’ personalized preferences and their trade-offs on nonfunctional attributes during service selection. The method allows users to represent their elastic nonfunctional requirements and associated importance using linguistic terms to specify their personalized trade-off strategies. We present examples showing how the service selection framework is used and a prototype with real-world airline services to evaluate the proposed frameworks application.

An Effective Dynamic Web Service Selection Strategy with Global Optimal QoS Based on Particle Swarm Optimization Algorithm

Dynamic Web service selection with global QoS optimization in Web service composition is a critical issue in Web service composition. In order to solve the problem, based on intelligent optimal theory of particle swarm optimization (PSO) algorithm, we propose a strategy PSO-GODSS (global optimization of dynamic Web service selection based on PSO) algorithm to implement Web service selection with QoS global optimization. The basic idea of the algorithm is to transform the original Web service selection problem into a multi-objective services composition optimization problem with global QoS constraints, which is further transformed into a single-objective problem by using the method of ideal point. Then, the theory of intelligent optimization of PSO is applied to produce a set of optimal services composition process with QoS constraints. Theoretical analysis and experimental results indicate the feasibility and efficiency of this algorithm, and the execution efficiency and convergence rate of PSO-GODSS are much better than that of multi-objective genetic algorithm used in prior work.

Service Selection Based on Personalized Preference and Trade-Offs among QoS Factors and Price

With the number of services with similar functionalities rising, service users place more emphasis on non-functional attributes of services such as quality of service (QoS) and price during service selection. However, previous QoS-driven service selection methods seldom consider relationships among non-functional requirements and do not allow explicit specification of personalized preferences and trade-offs from user perspectives to select services. In addition, most of them use weighted summation functions to aggregate multiple QoS factors to rank services for selection. This paper presents a novel service selection method, in which users can represent their non-functional requirements using linguistic terms and use their personalized trade-off strategies for service selection. The method employs fuzzy propositions in conjunction with QoS and price information to compute satisfaction degree of individual non-functional requirements. Based on the personalized trade-off strategy of a user, it then computes the overall satisfaction degree using fuzzy connective operators. The overall satisfaction degree of services is then used to rank services with equivalent functionalities and top-ranked services are recommended. An application example is presented to show the methods effectiveness.

Trust-Aware Service Recommendation via Exploiting Social Networks

With the rapid growth in the number of available services, recommending suitable services to users becomes increasingly important. A number of collaborative service recommendation methods based on user experiences have been proposed for this purpose. Most of them adopt the similarity-based Collaborative Filtering (CF) technique, which tends to identify similar users for a target user and recommends to the target user the services preferred by the similar users. However, a user similar to the target user is unnecessarily trustworthy to him/her. Therefore, the results recommended by similarity-based CF are probably unreliable. Moreover, existing service recommendation methods seldom incorporate social trust relationships among service users into service recommendation. In this paper, we propose a collaborative, trust-aware service recommendation method for service-oriented environments with social networks. The method is based on an integration of the user-service relation and the user-user social relation. Experimental results demonstrate that our service recommendation method significantly outperforms conventional similarity-based recommendation and trust-based service recommendation methods.

Collaborative Web Service Quality Prediction via Exploiting Matrix Factorization and Network Map

Quality of services (QoS) is an important concern in Web service recommendation or selection. Predicting QoS values of Web services based on their historical QoS records is an effective way to acquire Web service QoS, and thus has attracted considerable research interests. Recently, matrix factorization (MF), a well-known model-based collaborative filtering (CF) technique, has been successfully applied to the Web service QoS prediction. It is generally believed that MF can significantly outperform traditional memory-based CF techniques. However, previous work seldom considered the influence of the underlying network on Web service QoS when adopting MF for Web service QoS prediction. Hence, the prediction performance is not good enough. In this paper, we propose a network-aware Web service QoS prediction approach by integrating MF with the network map. By employing the network map, network distances between service users can be measured and neighborhoods of users are identified. Then, the traditional MF model is revamped by incorporating the constraint term that neighbor users are likely to perceive similar QoS of Web services. Experiments conducted on two real-world Web service datasets indicate that our approach outperforms previous MF and CF-based approaches in prediction accuracy.

P2P traffic optimization

Peer-to-peer (P2P) based content distribution systems have emerged as the main form for content distribution on the Internet, which can greatly reduce the distribution cost of content providers and improve the overall system scalability. However, the mismatch between the overlay and underlay networks causes large volume of redundant traffic, which intensifies the tension between P2P content providers and ISPs. Therefore, how to efficiently use network resources to reduce the traffic burden on the ISPs is crucial for the sustainable development of P2P systems. This paper surveys the state-of-art P2P traffic optimization technologies from three perspectives: P2P cache, locality-awareness and data scheduling. Technological details, comparison between these technologies and their applicabilities are presented, followed by a discussion of the issues that remain to be addressed and the direction of future content distribution research.