Requirements Modeling for Enterprise Applications of Machine Learning and Artificial Intelligence

点击翻译按钮获取中文摘要

In information systems engineering, requirements models play an important role in enabling business architects and stakeholders to conceive of innovative solutions to business problems, serving to define requirements for system development and implementation. Today, a new generation of IT systems is presenting fresh challenges for requirements engineering. These systems leverage the massive amounts of data from social media, sensor networks, mobile devices, as well as transactional data. Unlike traditional systems whose power derives from efficient and accurate execution of prescribed processes and transactions, these new software systems help generate insights and predictions and make recommendations using machine learning (ML) and artificial intelligence (AI). They have new capabilities and properties that are not well represented or analyzed through existing requirements models. Some examples of applications increasingly incorporating ML/AI include sales prediction, personalized marketing, diagnosis and recommendation in healthcare, and navigation guides.   The objective of the proposed research is to develop conceptual modeling techniques to support requirements engineering for enterprise applications of ML and AI. The modeling techniques will enable analysts to systemically explore and analyze alternative ways of positioning ML/AI functionalities within an organizational context (i.e., within a business process and the larger business and IT environment) so as to meet organizational and business goals.   The research will build upon and extend existing requirements modeling approaches, including process modeling, goal modeling, and agent modeling, and will explore cognitive modeling, to address the special needs of enterprise ML/AI. Research results will including novel modeling language features, notations, methods and techniques. A combination of analytical and empirical research methods will be used. Research results will be evaluated through industrial case studies.