An Investigation into the Application of Iterated-Local-Search for Combinatorial Optimisation

我的Final-Year-Project[1]调查了迭代局部搜索[2]（ILS）在试图解决组合优化问题（称为模块化问题[3]）时是否比随机突变爬坡（RMHC）更好或更差。通过基于图的聚类对软件系统的底层结构进行逆向工程的能力可以使工程师理解大型系统源代码中明显的关系。监视软件系统的结构和内部关系可以为项目管理提供关键信息，并间接影响维护和开发团队。尽管这种研究有好处，但软件系统的复杂性可能会基于诸如添加功能或系统是否开源等因素呈指数级增长。为了研究这个问题，基于Munch[4]开发了一个框架，其中为ILS和RMHC算法提供了一个称为模块依赖图（MDG）的软件系统图。我们的初步实验包括16个软件系统。实验结果表明，与目前的技术相比，盲降仪更加可靠和准确。

My Final-Year-Project[1] investigated whether Iterated Local Search[2] (ILS) is better or worse than Random Mutation Hill Climbing (RMHC) at attempting to solve the combinatorial optimisation problem known as the Modularisation Problem[3]. The ability to reverse-engineer the underlying structure of software systems through graph-based clustering can enable engineers to understand the relationships evident within the source code of large systems. Monitoring the structure and internal relationships of a software system overtime can provide pivotal information for project management with indirect implications on maintenance and development teams. Despite the benefits of such research, the complexity of a software system can exponentially increase based on factors such as added features or whether a system is open-sourced. To investigate this problem, a framework was developed based on Munch[4] in which a software system diagram known as a Module Dependency Graph (MDG) is provided to both an ILS and RMHC algorithm. Our preliminary experiments consisted of 16 software systems. The results of these experiments proved that ILS was more reliable and accurate compared to the current state of the art techniques.