PROJECT TITLE :

Category-Aware API Clustering and Distributed Recommendation for Automatic Mashup Creation

ABSTRACT:

Mashup has emeraged as a promising manner to allow developers to compose existed APIs (services) to make new or value-added services. With the speedy increasing range of services printed on the Net, service recommendation for automatic mashup creation gains a ton of momentum. Since mashup inherently needs services with completely different functions, the recommendation result should contain services from varied classes. However, most existing recommendation approaches only rank all candidate services during a single list, that has 2 deficiencies. 1st, ranking services without considering to that classes they belong might lead to meaningless service ranking and have an effect on the advice accuracy. Second, mashup developers aren't always clear about which service categories they need and services in which categories cooperate higher for mashup creation. While not explicitly recommending which service categories are relevant for mashup creation, it remains tough for mashup developers to select proper services in an exceedingly mixed ranking list, that lower the user friendliness of advice. To beat these deficiencies, a completely unique category-aware service clustering and distributed recommending method is proposed for automatic mashup creation. First, a Kmeans variant(vKmeans) methodology based mostly on topic model Latent Dirichlet Allocation is introduced for enhancing service categorization and providing a basis for recommendation. Second, on high of vKmeans, a service category relevance ranking (SCRR) model, that combines Machine Learning and Collaborative Filtering, is developed to decompose mashup needs and explicitly predict relevant service classes. Finally, a class-aware distributed service recommendation (CDSR) model, that relies on a distributed machine learning framework, is developed for predicting service ranking order among each category. Experiments on a true-world dataset have proved that the proposed approach not only gains vital improvement- at precision rate but also enhances the diversity of advice results.