Ph.D. Seminar – A. Abdullah

Join us Friday, May 27 at 10:00am in Reynolds 219 for the 2nd Ph.D. seminar by Ph.D. Candidate A. Abdullah.

Title

Towards a Comprehensive Web Service Recommendation Framework

Abstract

Web services (WS) are integrated software components that facilitate interoperable machine-to-machine interaction over a network. With the emergence of Cloud and Mobile Computing paradigms which essentially adopts web services as a means of management, companies worldwide are actively deploying web services within their business environments. Thus, designing effective Web service recommendation mechanisms is attracting more academical and industrial efforts to fulfill the increasing need for automation. Most of previous works were directed towards designing recommendation models either for individual or composite services separately, in addition to the lack of consideration to user preferences and dynamicity nature of service working environment. In individual service recommendation, traditional Neighborhood-based Collaborative Filtering (CF) models fail to capture the actual relationships between users or services due to data sparsity. In contrast, Random Walk algorithm, which has been categorized as a sparsity-tolerant recommendation approach, suffers from poor performance in terms of recommendation accuracy. We design a recommendation model that achieves high recommendation accuracy, through similarity integration, over the traditional Random Walk. First, we propose an Integrated-Model QoS-based Graph, in which weighted QoS magnitudes and User/Service similarity measurements are fused as Bipartite and Unipartite Graphs. In addition, Jaccard coefficient is introduced in several variants to separately compute similarities of both Users and Services. Then, Top-k Random Walk (RW) algorithm is applied to generate final recommendation list to active users. In composite service recommendation, the analogy between Service and Agent Computing paradigms suggests that incorporating both technologies will likely lead to a more effective hybrid service model. Therefore, we adopt an agent-based approach to Web Service Composition (WSC), in which Service Dependency Graph (SDG) is constructed as an AND/OR graph, distributed among the agent community members. Upon receiving a user composition request, agents perform internal reasoning and corporate through a communication protocol attempting to find a solution. A key feature of the model is the ability to apply user preferences within the composition process.

To demonstrate the effectiveness of our framework, experiments are conducted on public datasets. Analysis of the results shows significant improvement in terms of recommendation accuracy (with more tolerance to data sparsity) and communication cost.

Advisor: Dr. Xining Li

Advisory Committee: Dr. Stefano Gregory