REU Site: Software and Data Analytics

该项目将在东卡罗来纳大学(ECU)建立一个为期三年的软件和数据分析REU网站。它将在暑期学期为10名本科生提供为期10周的研究项目。师生互动以及学生之间的互动将采取不同的形式，包括每日Scrum会议、教程、每周会议、讲座、研讨会、小组会议和实地考察。REU项目将允许不同的本科生体验尖端研究。学生将获得宝贵的研究技能，为他们未来的研究领域做好准备，同时帮助他们发展成为自力更生的STEM研究人员。此外，他们对研究的接触将激励他们继续研究生学习。最后，REU项目将在暑期项目结束后为学生提供与他们的教师导师和全国各地的学生同行合作的机会。样本研究项目涵盖了软件和数据分析领域的公开研究课题。面向编程语言学习者的代码推荐研究机器学习技术，以构建针对初级程序员的代码推荐系统，同时考虑到他们的编程知识水平。智能程序更新检测和自动化使用软件系统的版本历史来了解与软件库的使用(通过应用程序编程接口或API)相关的代码如何演变，以识别何时需要发生这种演变，并构建转换脚本以部分或完全自动化支持较新的API版本所需的更改。人机协同对话系统探索了自动回归测试用例优先排序的技术，该技术利用了来自信息检索的技术，例如术语相似度。链接恢复系统研究如何使用信息检索技术来恢复程序要求、错误报告和项目源代码之间的可追溯性链接。使用机器学习来评估软件开发工作量探索了使用机器学习技术来评估软件开发工作量。了解隐式扩展API研究了机器学习对API推荐的使用，特别是在代码中隐式创建的动态语言API的上下文中。用于生物测定数据分析的机器学习算法使用机器学习技术和移动应用程序使用数据(例如，关于滑动手势)的组合来推断应用程序用户的人口统计特征。机器学习算法的性能评估探索了机器学习用于预测的使用，使用了加密货币的第二天收盘价的例子。参与这些项目的学生将学习包括代码推荐系统、静态程序分析、程序转换、机器学习中的经典分类技术(例如，k近邻)、深度学习、信息检索、软件测试、软件维护、软件仓库挖掘、软件质量度量、加密货币以及算法性能的理论和经验测量等主题。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

This project will establish a three-year REU site in software and data analytics at East Carolina University (ECU). It will offer a ten-week research program for ten undergraduate students during summer semesters. The faculty-student interaction, as well as interaction among students, will take different forms, including daily Scrum meetings, tutorials, weekly meetings, lectures, seminars, group meetings, and field trips. The REU project will allow a diverse pool of undergraduate students to experience cutting-edge research. Students will gain valuable research skills that will prepare them for their future fields of study, while helping them to develop into self-reliant STEM researchers. Furthermore, their exposure to research will motivate them to continue to graduate studies. Finally, the REU project will provide students with an opportunity to collaborate with their faculty mentors and student peers across the nation after the summer program ends. The sample research projects cover open research topics in software and data analytics. Code Recommendation for Programming Language Learners investigates machine learning techniques for building code recommendation systems aimed at beginning programmers, taking their level of programming knowledge into account. Intelligent Program Update Detection and Automation uses version histories of software systems to understand how code related to uses of a software library (via an Application Programming Interface, or API) evolves, to identify when this evolution needs to occur, and to build transformation scripts to partially or fully automate the changes needed to support a newer API version. Human-Computer Collaborative Dialogue Systems explores techniques for automated regression test case prioritization that utilizes techniques from information retrieval such as term similarity. Link Recovery Systems investigates the use of information retrieval techniques for recovering traceability links between program requirements, bug reports, and project source code. Using Machine Learning to Estimate Software Development Effort explores the use of machine learning techniques to estimate software development effort. Understanding Implicit Extension APIs investigates uses of machine learning for API recommendation, specifically in the context of APIs in dynamic languages that are created implicitly in the code. Machine Learning Algorithms for Biometric Data Analysis uses a combination of machine learning techniques and mobile application usage data (e.g., about swipe gestures) to infer demographic characteristics of app users. Performance Evaluation of Machine Learning Algorithms explores the use of machine learning for prediction, using the example of the next day closing price for crypt-currencies. Students participating in these projects will learn about topics including code recommendation systems, static program analysis, program transformation, classical techniques for classification in machine learning (e.g., k-nearest neighbors), deep learning, information retrieval, software testing, software maintenance, software repository mining, software quality metrics, crypto-currencies, and both theoretical and empirical measurements of algorithm performance.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.