Buqing Cao

An Efficient Search Strategy for Service Provider Selection in Complex Social Networks

The trustworthiness of service providers plays an important role when a consumer selects a service. This paper studies the problem of how to efficiently search and select trustworthy service providers for users in social networks consisting of service providers and consumers. A trust value between two participants can be derived by existing methods from the optimal trust path between them in a social network. When more than one trust factors are taken into consideration, the exact optimal trust path selection algorithm is NP-complete. Although several heuristic algorithms have been proposed to find approximate solutions, their time complexities are still too high to be acceptable in practice, especially when they are used in very large scale social networks. Focusing on reducing trust path searching time, this paper proposes an efficient preprocessing-based search strategy. It exploits structural properties of the social networks and builds an advanced data structure from preprocessing, which can be used to simplify and accelerate the trust path searching. Experimental results show our strategy is very efficient and nearly achieves a constant time complexity. The computed trustworthiness based on our method has excellent performance close to that of the best existing heuristic algorithm.

Integrated Content and Network-Based Service Clustering and Web APIs Recommendation for Mashup Development

The rapid growth in the number and diversity of Web APIs, coupled with the myriad of functionally similar Web APIs, makes it difficult to find most suitable Web APIs for users to accelerate and accomplish Mashup development. Even if the existing methods show improvements in Web APIs recommendation, it is still challenging to recommend Web APIs with high accuracy and good diversity. In this paper, we propose an integrated content and network-based service clustering and Web APIs recommendation method for Mashup development. This method, first develop a two-level topic model by using the relationship among Mashup services to mine the latent useful and novel topics for better service clustering accuracy. Moreover, based on the clustering results of Mashups, it designs a collaborative filtering (CF) based Web APIs recommendation algorithm. This algorithm, exploits the implicit co-invocation relationship between Web APIs inferred from the historical invocation history between Mashups clusters and the corresponding Web APIs, to recommend diverse Web APIs for each Mashups clusters. The method is expected to not only find much better matched Mashups with high accuracy, but also diversify the recommendation result of Web APIs with full coverage. Finally, based on a real-world dataset from ProgrammableWeb, we conduct a comprehensive evaluation to measure the performance of our method. Compared with existing methods, experimental results show that our method significantly improves the accuracy and diversity of recommendation results in terms of precision, recall, purity, entropy, DCG and HMD.

Effective Mashup Service Clustering Method by Exploiting LDA Topic Model from Multiple Data Sources

Mashup is emerging as a promising software development method for allowing software developers to compose existing Web APIs to create new or value-added composite Web services. However, the rapid growth in the number of available Mashup services makes it difficult for software developers to select a suitable Mashup service to satisfy their requirements. Even though clustering based Mashup discovery technique shows a promise of improving the quality of Mashup service discovery, Mashup service clustering with high accuracy for discovery of Mashup services is still a challenge problem. In this paper, we propose a novel Mashup service clustering method for Mashup service discovery with high accuracy by exploiting LDA topic model built from multiple data sources. It enables to infer topic probability distribution of Mashup services, which serves as a basis of computation of similarity of Mashup services. K-means and Agnes algorithm are used to perform Mashup service clustering in terms of their similarities. Compared with other service clustering approaches, experimental results show that our approach achieves significant improvement in terms of precision, recall and F-measure rate, which will improve Mashup service discovery.

WE-LDA: A Word Embeddings Augmented LDA Model for Web Services Clustering

Due to the rapid growth in both the number and diversity of Web services on the web, it becomes increasingly difficult for us to find the desired and appropriate Web services nowadays. Clustering Web services according to their functionalities becomes an efficient way to facilitate the Web services discovery as well as the services management. Existing methods for Web services clustering mostly focus on utilizing directly key features from WSDL documents, e.g., input/output parameters and keywords from description text. Probabilistic topic model Latent Dirichlet Allocation (LDA) is also adopted, which extracts latent topic features of WSDL documents to represent Web services, to improve the accuracy of Web services clustering. However, the power of the basic LDA model for clustering is limited to some extent. Some auxiliary features can be exploited to enhance the ability of LDA. Since the word vectors obtained by Word2vec is with higher quality than those obtained by LDA model, we propose, in this paper, an augmented LDA model (named WE-LDA) which leverages the high-quality word vectors to improve the performance of Web services clustering. In WE-LDA, the word vectors obtained by Word2vec are clustered into word clusters by K-means++ algorithm and these word clusters are incorporated to semi-supervise the LDA training process, which can elicit better distributed representations of Web services. A comprehensive experiment is conducted to validate the performance of the proposed method based on a ground truth dataset crawled from ProgrammableWeb. Compared with the state-of-the-art, our approach has an average improvement of 5.3% of the clustering accuracy with various metrics.

Service Selection Based on User Privacy Risk Evaluation

Requesting a service on the Internet may require the user’s privacy data, and thus raising the risk of the user’s privacy leakage and violation. Hence, it is necessary for users to select services that protect their privacy information. However, previous studies on service selection usually focused only on the quality of service, seldom had they considered the user’s privacy concern. As such, their results may be unable to meet the user’s privacy protection requirement. Aiming at reducing the privacy risk of users in service selection, this paper proposes a fuzzy logic service selection approach. The approach uses a fuzzy model to allow a service user specifying personalized privacy preference and a service provider specifying flexible privacy requirements; then it leverages the service’s reputation, privacy policy and the user’s privacy preference to compute the privacy risk for each service candidate; finally, it ranks all service candidates based on their privacy risk degrees. Examples and evaluations show that the proposed approach is effective and efficient for reducing privacy risk in service selection.

Web Page Recognition Algorithm Based on Link Analysis in Theme Search Engine

Web page recognition is a problem in the design of web crawler in theme search engine. This paper designs a web page recognition algorithm based on link analysis to solve this problem. The main idea of this algorithm is to get the relevant web page recognition model through a combination of link analysis and theme URL knowledge base, based on the idea of statistics and social network analysis. Through the experiment, the precision rate of this algorithm is over 93 percent, and the recall rate is up to 85.4 percent. So the experiment is significant, better than other web page recognition algorithm. Experimental results show the feasibility and effectiveness of this algorithm.

Mashup Service Classification and Recommendation Based on Similarity Computing

Because of the excellent performance of Mashup service in the service composition, Mashup service is used more and more. It is meaningful for service management, discovery and composition that how to achieve effective Mashup service classification and recommendation. We analyze the service network consisted of Mashup applications, Web API services and Tag functions, basing on the rule that there are connections among those Mashups if some Mashups call the same APIs and are marked by the same Tags, and the degree of the connection can be described by similarity, and build 13 kinds of networks and visualize them. Based on built service network, this paper proposes an automatic service classification algorithm that each connected sub-graph is justly a classification in the network consisted of a same kind of service node, and a service recommendation method based on the similarity sorting. We use the Web API data crawled from ProgrammableWeb. The result of our experiment shows the composite index of precision rate and recall rate is up to 87.44%.

Integrating Tag, Topic, Co-Occurrence, and Popularity to Recommend Web APIs for Mashup Creation

With the rapid development of Web APIs, selection of the suitable Web APIs from the service repositories for users to build Mashup applications becomes more and more difficult. Even if the existing methods show significant improvements in Web API recommendation, it is still challenging to recommend similar, diverse, and relevant Web APIs with high accuracy. In this paper, we propose a novel Web API recommendation method, which integrates tag, topic, co-occurrence, and popularity factors to recommend Web APIs for Mashup creation. This method, firstly exploits the enriched tags and topics information of Mashups and Web APIs derived by the relational topic model to calculate the similarity between Web APIs and the similarity between Mashups. Secondly, it uses the invocation times and category information of Web APIs to derive their popularity. Thirdly, multi-dimensional information, such as similar Mashups, similar Web APIs, co-occurrence and popularity of Web APIs, are modeled by factorization machines to predict and recommend top-k similar, diverse, relevant Web APIs for a target Mashup. Finally, we conduct a set of experiments, and experimental results show that our approach achieves a significant improvement in terms of precision, recall, F-measure, compared with other existing methods.

Web API Recommendation for Mashup Development Using Matrix Factorization on Integrated Content and Network-Based Service Clustering

Finding appropriate web APIs to develop mashup services is becoming difficult because of increasing number of web APIs offered from different sources. If we can recommend relevant web APIs for a mashup service based on its requirements, it will help software developers to find suitable APIs easily instead of searching from thousands of web APIs. Although there are many existing methods to recommend web APIs for mashup services, their recommendation accuracies and diversities are still not high. We will present a novel approach in this paper to produce better web API recommendation results in terms of accuracy and diversity. It is a matrix factorization based API recommendation method for Mashup services. It uses a two-level topic model for clustering Mashup services. We used a dataset from programmableWeb to perform experiments and compared results of our method with other existing methods. Its evaluation results show that our matrix factorization based recommendation archives better API recommendation accuracy and diversity for Mashup services.

Multi-relation Based Manifold Ranking Algorithm for API Recommendation

The number of APIs on the Web has increased rapidly in recent years. It becomes quite popular for developers to combine different APIs to build innovative Mashup applications. However, it is challenging to discover the appropriate ones from enormous APIs for Mashup developers (i.e., API users). In order to recommend a set of APIs that most satisfy the users’ requirements, we propose a multi-relation based manifold ranking approach. The approach exploits the textual descriptions of existing Mashups and APIs, as well as their composition relationships. It firstly groups Mashups into different clusters according to their textual descriptions, then explores multiple relations between Mashup clusters and between APIs. Finally, it employs a manifold ranking algorithm to recommend appropriate APIs to the user. Experiments on a real-world dataset crawled from ProgrammableWeb.com validate the effectiveness of the proposed approach.