Event-based data are encountered daily in many disciplines and are used for various purposes. They are collections of ordered sequences of events where each event has a start time and a duration. Examples of such data include medical records, internet surfing records, transaction records, industrial process or system control records, and activity diary data. This thesis is concerned with the exploration of event-based data, and in particular the identification and analysis of sequences within them. Sequences are interesting in this context since they enable the understanding of the evolving character of event data records over time. They can reveal trends, relationships and similarities across the data, allow for comparisons to be made within and between the records, and can also help predict forthcoming events.The presented work has researched methods for identifying and exploring such event-sequences which are based on modern visualization, interaction and data mining techniques. An interactive visualization environment that facilitates analysis and exploration of event-based data has been designed and developed, which permits a user to freely explore different aspects of this data and visually identify interesting features and trends. Visual data mining methods have been developed within this environment, that facilitate the automatic identification and exploration of interesting sequences as patterns. The first method makes use of a sequence mining algorithm that identifies sequences of events as patterns, in an iterative fashion, according to certain user-defined constraints. The resulting patterns can then be displayed and interactively explored by the user.The second method has been inspired by web-mining algorithms and the use of graph similarity. A tree-inspired visual exploration environment has been developed that allows a user to systematically and interactively explore interesting event-sequences.Having identified interesting sequences as patterns it becomes interesting to further explore how these are incorporated across the data and classify the records based on the similarities in the way these sequences are manifested within them. In the final method developed in this work, a set of similarity metrics has been identified for characterizing event-sequences, which are then used within a clustering algorithm in order to find similarly behavinggroups. The resulting clusters, as well as attributes of the clusteringparameters and data records, are displayed in a set of linked views allowing the user to interactively explore relationships within these. The research has been focused on the exploration of activity diary data for the study of individuals' time-use and has resulted in a powerful research tool facilitating understanding and thorough analysis of the complexity of everyday life.

Predictive maintenance plays a major role in operational cost reduction in several industries and the railway industry is no exception. Predictive maintenance relies on real time data to predict and diagnose technical failures. Sensor data is usually utilized for this purpose, however it might not always be available. Events data are a potential substitute as a source of information which could be used to diagnose and predict failures. This thesis investigates the use of events data in the railway industry for failure diagnosis and prediction. The proposed approach turns this problem into a sequence classification task, where the data is transformed into a set of sequences which are used to train the machine learning algorithm. Long Short-Term Memory neural network is used as it has been successfully used in the past for sequence classification tasks. The prediction model is able to achieve high training accuracy, but it is at the moment unable to generalize the patterns and apply them on new sets of data. At the end of the thesis, the approach is evaluated and future steps are proposed to improve failure diagnosis and prediction.

Low-power wireless networks consist of spatially distributed, resource-constrained devices – also referred to as nodes – that are typically equipped with integrated or external sensors and actuators. Nodes communicate with each other using wireless transceivers, and thus, relay data – e. g., collected sensor values or commands for actuators – cooperatively through the network. This way, low-power wireless networks can support a plethora of different applications, including, e. g., monitoring the air quality in urban areas or controlling the heating, ventilation and cooling of large buildings. The use of wireless communication in such monitoring and actuating applications allows for a higher flexibility and ease of deployment – and thus, overall lower costs – compared to wired solutions. However, wireless communication is notoriously error-prone. Message losses happen often and unpredictably, making it challenging to support applications requiring both high reliability and low latency. Highly reliable, low-latency communication – along with high energy-efficiency – are, however, key requirements to support several important application scenarios and most notably the open-/closed-loop control functions found in e. g., industry and factory automation applications. Communication protocols that rely on synchronous transmissions have been shown to be able to overcome this limitation. These protocols depart from traditional single-link transmissions and do not attempt to avoid concurrent transmissions from different nodes to prevent collisions. On the contrary, they make nodes send the same message at the same time over several paths. Phenomena like constructive interference and capture then ensure that messages are received correctly with high probability. While many approaches relying on synchronous transmissions have been presented in the literature, two important aspects received only little consideration: (i) reliable operation in harsh environments and (ii) support for event-based data traffic. This thesis addresses these two open challenges and proposes novel communication protocols to overcome them.

This thesis investigates the application of knowledge transfer techniques to process event-based data forperson detection in area surveillance. A teacher-student model setup is employed, where both modelsare pretrained on conventional visual data. The teacher model processes visual images to generate targetlabels for the student model trained on event-based data, forming the baseline model. Building onthis, the project incorporates feature-based knowledge transfer, specifically transferring features fromthe Feature Pyramid Network (FPN) component of the Faster R-CNN ResNet-50 FPN network. Resultsindicate that response-based knowledge transfer can effectively finetune models for event-based data.However, feature-based knowledge transfer yields mixed results, requiring more refined techniques forconsistent improvement. The study identifies limitations, including the need for a more diverse dataset,improved preprocessing methods, labeling techniques, and refined feature-based knowledge transfermethods. This research bridges the gap between conventional object detection methods and event-baseddata, enhancing the applicability of event cameras in surveillance applications.