Ci-Wei Lan’s research while affiliated with IBM and other places

Content-based recommendation techniques usually require a large number of training examples for model construction, which however may not always be available in many real-world scenarios. To address the training data availability constraint common to the content-based approach, we develop a collaborative expansion-based approach to expand the size of training examples, which could lead to improved content-based recommendations. We use a book rating data set collected from Amazon to evaluate our proposed method and compare its performance against those of two salient benchmark techniques. The results show that our method outperforms the benchmark techniques consistently and significantly. Our method expands the size of training examples for a focal customer by leveraging the available preferences of his or her referent group, and thereby better supports personalized recommendations than existing techniques that solely follow content-based or collaborative filtering, without incurring costs to identify, collect, and analyze additional information. This study reveals the value and feasibility of collaborative expansion as a viable means to increase training size for the focal customer and thus address the training data availability constraint that seriously hinders the performance of content-based recommender systems.

The microservice architecture (MSA) differs fundamentally from the monolithic, layered architecture. The use of microservices provides a high degree of autonomy, composability, scalability, and fault-tolerance. MSA is regarded by many as a promising architecture for smart-city applications; however, a number of issues remain, including (1) the management of complex call relationships among microservices; (2) ensuring the quality of the overall software system even as new microservices are added and existing ones are modified, and (3) locating existing microservices that satisfy new requirements. In this paper, we propose a novel approach to the development of microservice-based systems, referred to as GSMART (Graph-based and Scenario-driven Microservice Analysis, Retrieval and Testing). GSMART enables the automatic generation of a “Service Dependency Graph (SDG)” by which to visualize and analyze dependency relationships between microservices as well as between services and scenarios. It also enables the automatic retrieval of test cases required for system changes to reduce the time and costs associated with regression testing. A microservice retrieval method using VSM and word2vec accelerates the development of new microservices tailored specifically to the needs of users based on user-provided scenarios. Experiment results demonstrate the feasibility, effectiveness, and efficiency of all of the main features of GSMART.

Web API is the lightweight version to the SOAP (Simple Object Access Protocol) service and usually applies REST (Representational State Transfer) as the architectural style. Nowadays, the service-oriented computing paradigm is shifting from the SOAP services to the Web APIs, i.e., RESTful services. Rather than hard coding from scratch, the methodology of service-oriented computing emphasizes identifying the granularity services and composing services into executable applications. Quality of Service (QoS) is always an important concern in the selection of services. The QoS driven service selection problem for service composition has been proven as NP-Complete since there are nm combinations for a given composite service involving m tasks with n candidate services for each task. To address this issue, this study proposes a methodology based on Genetic Algorithms for the selection of Web APIs. The proposed approach in this research provide several features: 1) e-Pareto dominance relations are introduced to help the discrimination of composite QoS, 2) the e value for each dimension is determined by the AHP-based method (AHP: Analytic Hierarchy Process), and 3) the Genetic Algorithm is employed to find out the optimal combinations of Web APIs. Experimental results demonstrate that the proposed approach can effectively locate the combination of Web APIs based on the user-preferred QoS indicators without sacrificing other kinds of service quality.

A method and a system for preserving privacy of a dataset are provided. In the method, a k-anonymity value with respect to a sensitive data field is determined according to at least one first quasi-identifier. Data entries in each group have the same value in the one or more fields of the first quasi-identifier and data entries in different groups have different values in the one or more fields of the first quasi-identifier. A first group and a second group among the plurality of groups are determined according to the reference number Kr, where the first group and the second group are merged into a merging group. The number of data entries in the merging group is not less than a reference number Kr. One or more fields of at least one first quasi-identifier is masked for the merging group.

Web service discovery is the process of locating web services to satisfy the requirements of service requesters, and as such, plays an important role to realize business-to-business interoperability. Today, many service discovery mechanisms are available, roughly classified into two categories: text-based service retrieval and ontology-based service matching. However, both kinds of service discovery approaches have their limitations in finding out suitable web services for users. Text-based approaches cannot reflect the semantics of the service capability. Ontology-based approaches are likely infeasible and non-scalable since service providers need to spend considerable efforts to annotate the service descriptions. This study proposes an approach to searching for web services, called LS3 (Lexical and Semantic Service Search) architecture. This proposed approach includes two main sub processes: query expansion and service ranking to enable the retrieval of relevant web services during the discovery process by considering lexical similarity and semantic similarity.

Enterprise asset management is a system which manages all kinds of reusable resources in an enterprise no matter what departments or business units produce. The efficiency and effectiveness of asset management will significantly impact on development cost and enterprise competiveness. In this paper, we address several challenges including lacking of asset description model, non-solution reuse pattern and isolated assets. By identifying key issues, we propose solution ideas to provide a robust asset description framework so that a lot of asset types have their corresponding attributes. Furthermore, we also propose a graph match solution to take topological structures in user query into consideration for solution asset matchmaking.

Secret sharing is widely used in secret image protection since 1979. Image secret sharing schemes are used to encrypt the secret image into several meaningless share images. Later on, the secret image can be reconstructed while collecting the share images. The authentication mechanism is a way to verify whether the reconstructed secret image is authentic. The general authentication might increase the cost of key maintenance. To reduce the cost, this paper proposed a self-authentication mechanism for image secret sharing scheme. In addition, the reconstructed image can be completely reconstructed. The experiments give the positive data to show the feasibility of the proposed scheme.