The rise of Artificial Intelligence (AI) enables enterprises to manage large amounts of data in order to derive predictions about future performance and to gain meaningful insights. In this context, descriptive and predictive analytics has gained a significant research attention; however, prescriptive analytics has just started to emerge as the next step towards increasing data analytics maturity and leading to optimized decision making ahead of time. Although machine learning for decision making has been identified as one of the most important applications of AI, up to now, prescriptive analytics is mainly addressed with domain-specific optimization models. On the other hand, existing literature lacks generalized prescriptive analytics models capable of being dynamically adapted according to the human preferences. Reinforcement Learning, as the third machine learning paradigm alongside supervised learning and unsupervised learning, has the potential to deal with the dynamic, uncertain and time-variant environments, the huge states space of sequential decision making processes, as well as the incomplete knowledge. In this paper, we propose a human-augmented prescriptive analytics approach using Interactive Multi-Objective Reinforcement Learning (IMORL) in order to cope with the complexity of real-life environments and the need for optimized human-machine collaborations. The decision making process is modelled in a generalized way in order to assure scalability and applicability in a wide range of problems and applications. We deployed the proposed approach in a stock market case study in order to evaluate the proactive trading decisions that will lead to the maximum return and the minimum risk that the user’s experience and the available data can yield in combination.
Business analytics use advanced techniques that can analyze and process large and diverse data sets in order to generate valuable insights and lead to better business decisions. Of the three types of business analytics – descriptive, predictive, and prescriptive – only the latter focus on decision making. This paper aims to address two limitations of existing approaches in prescriptive analytics: (i) the lack of a transparent integration between predictive and prescriptive analytics and (ii) the incorporation of human knowledge and experience within the decision-making process. In order to address these points, the paper develops a framework that integrates data-driven predictions and the decision-making process by taking account human experience. The framework adopts interactive reinforcement learning algorithms and provides a concrete approach for data-driven human-AI collaboration. The main challenges and limitations of the approach are also discussed.
Augmented analytics is an emerging topic which deals with the enhancement of analytics with conversational interfaces as well as the exploitation of the human knowledge representation through intelligent digital assistants allowing users to easily interact with data and insights. The communication with the user by voice poses new challenges to the development and execution of data analytics services. In this paper, we outline a framework for implementing quality analytics for decision augmentation through optimized human-AI interaction. Our approach aims to reduce the number of quality issues through fast, mobile, and easy access to quality predictions for products and processes. An application case is the production of white goods is presented.
Due to the emergence of sensing technology, a large number of sensors is used to monitor the health state of manufacturing equipment, thus enhancing the capabilities of predicting abnormal behaviours in (near) real-time. However, existing algorithms in predictive maintenance suffer from several limitations related to their scalability, efficiency, and reliability preventing their wide application to various industries. This paper proposes an approach for real-time prediction of the equipment health state using time-domain features extraction, Long Short-Term Memory (LSTM) Neural Networks, and Bayesian Online Changepoint Detection (BOCD). The proposed approach is applied to a real-life case in the steel industry and extensive experiments are performed. The paper also discusses the results and the conclusions drawn from the proposed approach.
Voice-enabled assistants, such as Alexa and Google Assistant, are among the fastest-growing information technology applications. Their technological foundation matured over the last years and reached a point where new application areas in challenging business environments become a certainty. Maintenance in manufacturing is one of these areas. This paper presents expectations, requirements, and a concept for a voice-enabled digital intelligent assistant that supports maintenance activities. We identified process monitoring, task execution, reporting, problem-solving, and maintenance planning as the key functional modules for an assistant. Realizing them depends on basic, utility, and maintenance functions. Our discussion states that all fundamental technologies and tools to realize an assistant for maintenance exist, but they have constraints. For instance, Speech-to-Text mechanisms lack transparent and performant solutions, and natural language understanding must rely on small datasets, which is challenging. We argue that continuous improvement and systematic evaluation of an assistant prototype is important to create high-quality training data. Trial-and-error is common because some technologies still mature, and conversation designers lack design patterns for the maintenance domain. Challenges for system adoption include providing an outstanding user experience, handling factory-specific jargon, and the limited availability of easy-to-use data exchange interfaces for machines and business applications. We conclude that further efforts on interoperability, technology stack management, AI-focused change management, and education programs are necessary. Furthermore, the accountability of AI systems is a cost factor for the assistant’s service providers and the client companies in manufacturing – AI insurance services, human-in-the-loop functions, user training, and professional education are actions to address this issue.