Pattern Discovery And Optimization Technologies In Web Mashup

In the past decades, web-related technologies have been developed rapidly, which makes Web applications be applied widely, especially in wireless environment. Open API, often offered by Web companies, packages Web services, content, data and other resources into a series of data interfaces which can be accessed and identified easily on the Web. The third-party developers can use Open APIs to build new Web applications, which we call as Mashup applications. Generally, Web Mashup refers to mash two or more different Open APIs into a certain Web application in order to fulfil a specific task. For instance, HousingMaps.com is a typical Mashup application which is formed by the Google Map API and Craigslist’s RSS. Since it is published on-line, Mashup technology gets widespread concern in both academic and industrial fields. In order to provide best services to their clients, Web organizations have published their core business services as Open APIs, which leads to the sharp increment of number and varieties of Open APIs in recent years. The rich Open APIs gives the realization foundation for developing Mashup applications. However, with a large number of Open APIs and Mashups, selecting appropriate Open APIs, Open API orchestration patterns for simplifying Mashup building process becomes a big chanllenge for developers. By taking the above problem into consideration, this work arches fundmental Mashup modes for Web application. In detail, the main contributions of this thesis are listed as follows:(1) Proposing Mashup directed orchestration model. A Mashup application integrates data or sevices from two or more Open APIs. It is assumed that Open API is the smallest element of Mashup. The data exchange among Open APIs forms the combination relationship between Open APIs. There are two kinds of relationship between Open APIs: serial combination and parallel combination. We present Mashup directed orchestration model to describe the combination of Open APIs formally.(2) Proposing frequent Mashup sub-orchestration mining algorithm. Mashup directed orchestration model is a directed graph whose nodes refer to Open APIs, and directed edge signifies the relationship between Mashup’s Open APIs. An obvious feature of the model is that it has few nodes and edges. In this work, each Open API is given a global identification number. Combining all above charater of Mashup, a frequent sub-orchesration pattern discovery algorithm is proposed which can compute the frequent patterns in a more accurate and effective way(3) Proposing MPLIndex based on frequent Mashup pattern. Frequent pattern often used to describe Mashup mode characteristics. Taking advantage of the theory of form concepts, the Mashup patterns are expressed as concept’s extension and the characteristics extracted from frequent patterns are represented as intension. Based on this, we propose a new index named as MPLIndex. MPLIndex can be cached to improve the performance of searching desired Mashup patterns and the orchestration of Mashup.(4) Demonstrating the algorithm FSOMM and MPLIndex with experiments. The performance of the algorithm is verified by real and simulated data. The experiments show that by applying FSOMM algorithm on real data sets, we successfully discover the frequent orchestration patterns of real Mashup applications. The extensibility and query performance of MPLIndex are verified by simulated data.