Deep Learning for Process and Business Analytics

Modern businesses gather tremendous amounts of data on their operations. These businesses often operate in a complex environment, therefore, the recorded data often comes from numerous varied sources and is stored in various forms. Unsurprisingly, one of the main challenges of today's businesses is to extract valuable insights from this plethora of recorded data. Consequently, there has been a growing interest in the field of business analytics in recent decades. Generally, this field can be considered the science of discovering insights using analytical methods to support and improve decision-making within businesses. A brief introduction to business analytics will be provided in the first part of this dissertation. An overview of the structure of the dissertation is also provided in this part, in addition to discussing contributions made to two research disciplines which fall within the general domain of business analytics, namely predictive process monitoring and credit scoring. Process analytics can be used to identify and assess issues and opportunities for improving business processes. In recent decades, the field of process mining has emerged, which focuses on extracting insights from event logs, recorded and stored by information systems used to support business process executions. More recently, this field has increasingly focused on the development of predictive process monitoring techniques which can be used to provide real-time decision support during business process execution. Predictive process monitoring is the topic of the second part of this dissertation. First, an introduction to this field is given in addition to a development cycle for predictive process monitoring applications. Second, a direct data aware LSTM architecture is proposed for predicting the future trajectory and runtime of ongoing process executions. Third, a multilocation load state inter-case encoding framework is proposed and evaluated for remaining trace and runtime prediction using LSTM neural networks. Finally, the applicability of predictive process monitoring for supporting luggage handling operations at a large international airport is investigated. Since the mid-20th century, research in business analytics has increasingly focused on developing mathematical models to support loan approval processes. As a result, the field of credit scoring has emerged. The third part of this dissertation investigates the appropriateness of deep learning algorithms for credit sco