A Software Component Recommendation System Based on Development Context

在现代软件开发中，许多软件产品都是通过组合现有的软件组件，并只编写最小的所需功能来构建的。然而，从众多组件中选择必要的组件并不容易。例如，Java 2 SDK标准版（J2SE） Version 1.41_02被软件开发人员广泛使用，并为他们提供了5568个软件组件（基类）。从我们的试点实验中，我们发现在典型的软件开发中只有4%的Java 2 SDK软件组件被使用。这个结果表明仍然需要额外的编码，而目前有各种各样的软件组件可供重用。我们研究的目标是为基于组件的软件开发实现一个高效的环境。为了实现这个目标，在这个研究项目中，我们构建了一个系统，根据软件开发的上下文向开发人员推荐所需的软件组件。我们特别注重以下基本技术的开发：(A)。胎记：一种分析开发环境的技术，这意味着开发的进度和/或完整性，软件本身的特性和/或结构，等等，通过从正在开发的目标软件中提取特征的数量，(B)。协同过滤：一种识别和推荐具有与目标软件相似上下文的软件组件的技术，以及(C)。关联可视化：一种可视化技术，为开发人员提供系统向开发人员推荐软件组件的原因。我们将这三种技术整合到一个系统中，并对其进行了评估。

In modern software development, many of software products are built by combining existing software components and coding only minimal required features. However, selecting necessary components from a number of components is not easy. For instance, Java 2 SDK Standard Edition (J2SE) Version 1.41_02 is widely used by software developers and provides them with 5568 software components (basis classes). From our pilot experiments, we have found that only 4% of the software components of Java 2 SDK were used in typical software development. This result indicates additional coding is still much required, while a variety of software components are currently available to reuse.The goal of our study is to realize an efficient environment for components-based software development. Toward achieving the goal, in this research project, we constructed a system that recommends required software components to developers according to the context of software development. Especially we focused on developing the following elemental technologies:(A). Birthmark: a technology to analyze the development context which means the progress and/or completeness of development, features and/or structures of software itself, and so forth, by extracting the amount of characteristics from target software under development,(B). Collaborative Filtering: a technology to identity and recommend software components which have context similar to target software, and(C). Association Visualization: a visualization technology to provide developers with reasons why the system recommends software components to developers.We integrated the three technologies into a system and evaluated it.