Many machine learning models deployed on smart or edge devices experience a phase where there is a drop in their performance due to the arrival of data from new domains. This paper proposes a novel unsupervised domain adaptation algorithm called DIBCA++ to deal with such situations. The algorithm uses only the clusters’ mean, standard deviation, and size, which makes the proposed algorithm modest in terms of the required storage and computation. The study also presents the explainability aspect of the algorithm. DIBCA++ is compared with its predecessor, DIBCA, and its applicability and performance are studied and evaluated in two real-world scenarios. One is coping with the Global Navigation Satellite System activation problem from the smart logistics domain, while the other identifies different activities a person performs and deals with a human activity recognition task. Both scenarios involve time series data phenomena, i.e., DIBCA++ also contributes towards addressing the current gap regarding domain adaptation solutions for time series data. Based on the experimental results, DIBCA++ has improved performance compared to DIBCA. The DIBCA++ has performed better in all human activity recognition task experiments and 82.5% of experimental scenarios on the smart logistics use case. The results also showcase the need and benefit of personalizing the models using DIBCA++, along with the ability to transfer new knowledge between domains, leading to improved performance. The adapted source and target models have performed better in 70% and 80% of cases in an experimental scenario conducted on smart logistics. 

A large amount of data is generated from fields like IoT, smart monitoring applications, etc., raising demand for suitable data analysis and mining techniques. Data produced through such systems have many distinct characteristics, like continuous generation, evolving nature, multi-source origin, and heterogeneity, and in addition are usually not annotated. Clustering is an unsupervised learning technique used to group and analyze unlabeled data. Conventional clustering algorithms are unsuitable for dealing with data with the mentioned characteristics due to memory, computational constraints, and their inability to handle the heterogeneous and evolving nature of the data. Therefore, novel clustering approaches are needed to analyze and interpret such challenging data. 

This thesis focuses on building and studying advanced clustering algorithms that can address the main challenges of today's real-world data: evolving and heterogeneous nature. An evolving clustering approach capable of continuously updating the generated clustering solution in the presence of new data is initially proposed, which is later extended to address the challenges of multi-view data applications. Multi-view or multi-source data presents the studied phenomenon or system from different perspectives (views) and can reveal interesting knowledge that is invisible when only one view is considered and analyzed. This has motivated us to continue exploring data from different perspectives in several other studies of this thesis. Domain shift is another common problem when data is obtained from various devices or locations, leading to a drop in the performance of machine learning models if they are not adapted to the current domain (device, location, etc.). The thesis explores the domain adaptation problem in a resource-constraint way using cluster integration techniques. A new hybrid clustering technique for analyzing the heterogeneous data is also proposed. It produces homogeneous groups, facilitating continuous monitoring and fault detection.

The algorithms and techniques proposed in this thesis are evaluated on various data sets, including real-world data from industrial partners in domains like smart building systems, smart logistics, and performance monitoring of industrial assets. The obtained results demonstrated the robustness of the algorithms for modeling, analyzing, and mining evolving data streams and/or heterogeneous data. They can adequately adapt single and multi-view clustering models by continuously integrating newly arriving data.

In this study, we propose to explore multi-layered clustering to provide a context-aware data analytics tool for monitoring the network behavior of subsystems, such as a district heating (DH) network. Multi-layer clustering, in contrast to multi-view clustering, does not assume conditional independence of layers. The main idea of our approach is based on the integration of clustering models produced by considering different perspectives that capture information about the monitored subsystems' operational behavior or performance as well as their contextual environment. The initial clustering layer can reflect a static context, which is important for the subsystems' performance. It will be used as a base on which clustering models produced with respect to other analyzed operational characteristics and contexts will be layered. This will facilitate analysis and comparison of the subsystems' behavior in two comparable time periods and, eventually, identification of deviations that need attention. The proposed approach is evaluated and validated in a use case from the DH domain. The multi-layered clustering is applied and demonstrated to be robust for continuous context-aware analysis of the performance of a network of DH substations. 

Many industrial scenarios are concerned with the exploration of high-dimensional heterogeneous data sets originating from diverse sources and often incomplete, i.e., containing a substantial amount of missing values. This paper proposes a novel unsupervised method that efficiently facilitates the exploration and analysis of such data sets. The methodology combines in an exploratory workflow multi-layer data analysis with shared nearest neighbor similarity and hypergraph clustering. It produces overlapping homogeneous clusters, i.e., assuming that the assets within each cluster exhibit comparable behavior. The latter can be used for computing relevant KPIs per cluster for the purpose of performance analysis and comparison. More concretely, such KPIs have the potential to aid domain experts in monitoring and understanding asset performance and, subsequently, enable the identification of outliers and the timely detection of performance degradation.

Online instruction has become increasingly common as an alternative to face-to-face instruction (Crawford-Ferre & Wiest, 2012; Maertens et al., 2016; Ananga & Biney, 2017). One benefit with online instruction is that it is more easily accessible for students who are not able to fully access the more traditional face-to-face instruction on campus. After the Covid-19 pandemic, online instruction has gained further ground (Zhu & Liu, 2020; Kerres & Buchner, 2022; Li et al., 2022). At the same time, we have seen a rapid increase in new educational technologies, including that of virtual reality (Ding & Li, 2022; Al-Ansi et al., 2023; Zhang et al., 2022). Studies show that virtual reality (VR) can make the learning process more engaging and interactive (Jackson & Fagan, 2000; Ardiny& Khanmirza, 2018; Roopa et al., 2021) and that it can increase reception levels and train collaborative skills (Isik-Ercan et al., 2010; Petersen et al., 2023). This paper raises the question of how the use of virtual reality in online instruction affects learning experiences. While the participants in the pilot study displayed a genuine enthusiasm for using VR in an online setting, results showed a lack of knowledge in how to use VR to improve student learning. One area of investigation was concentration. Here, results were inconclusive as 50 % of the participants in group 1 (G1) were unsure of whether VR improves concentration, while 50 % of the participants in group 2 (G2) claimed that the use of VR does improve their concentration level. Another area of investigation was understanding the topic. The participants from G1 gave higher ratings than those who performed the experiment in G2, which implies that the impact was not as great as expected. In fact, the participants in G2 found that the VR equipment shifted focus from learning to other details in the visual medium. Another area was interactivity. Here, results indicated that VR technology has the didactic potential of engaging students and making them more interactive in the learning situation. The study concludes that while VR technology has the possibility of enhancing learning, a prerequisite is that both students and teachers have the skills and knowledge of how to use VR technology in a pedagogical setting; furthermore, a few technical modifications to the device itself are required.

Today’s smart monitoring applications need machine learning models and data mining algorithms that are capable of analysing and mining the temporal component of data streams. These models and algorithms also ought to take into account the multi-source nature of the sensor data by being able to conduct multi-view analysis. In this study, we address these challenges by introducing a novel multi-view data stream clustering approach, entitled MST-MVS clustering, that can be applied in different smart monitoring applications for continuous pattern mining and data labelling. This proposed approach is based on the Minimum Spanning Tree (MST) clustering algorithm. This algorithm is applied for parallel building of local clustering models on different views in each chunk of data. The MST-MVS clustering transfers knowledge learnt in the current data chunk to the next chunk in the form of artificial nodes used by the MST clustering algorithm. These artificial nodes are identified by analyzing multi-view patterns extracted at each data chunk in the form of an integrated (global) clustering model. We further show how the extracted patterns can be used for post-labelling of the chunk’s data by introducing a dedicated labelling technique, entitled Pattern-labelling. We study and evaluate the MST-MVS clustering algorithm under different experimental scenarios on synthetic and real-world data. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.

Domain shift is a common problem in many real-world applications using machine learning models. Most of the existing solutions are based on supervised and deep-learning models. This paper proposes a novel clustering algorithm capable of producing an adapted and/or integrated clustering model for the considered domains. Source and target domains are represented by clustering models such that each cluster of a domain models a specific scenario of the studied phenomenon by defining a range of allowable values for each attribute in a given data vector. The proposed domain integration algorithm works in two steps: (i) cross-labeling and (ii) integration. Initially, each clustering model is crossly applied to label the cluster representatives of the other model. These labels are used to determine the correlations between the two models to identify the common clusters for both domains, which must be integrated within the second step. Different features of the proposed algorithm are studied and evaluated on a publicly available human activity recognition (HAR) data set and real-world data from a smart logistics use case provided by an industrial partner. The experiment's goal on the HAR data set is to showcase the algorithm's potential in automatic data labeling. While the conducted experiments on the smart logistics use case evaluate and compare the performance of the integrated and two adapted models in different domains. © 2022 IEEE.

Data available today in smart monitoring applications such as smart buildings, machine health monitoring, smart healthcare, etc., is not centralized and usually supplied by a number of different devices (sensors, mobile devices and edge nodes). Due to which the data has a heterogeneous nature and provides different perspectives (views) about the studied phenomenon. This makes the monitoring task very challenging, requiring machine learning and data mining models that are not only able to continuously integrate and analyze multi-view streaming data, but also are capable of adapting to concept drift scenarios of newly arriving data. This study presents a multi-view clustering approach that can be applied for monitoring and analysis of streaming data scenarios. The approach allows for parallel monitoring of the individual view clustering models and mining view correlations in the integrated (global) clustering models. The global model built at each data chunk is a formal concept lattice generated by a formal context consisting of closed patterns representing the most typical correlations among the views. The proposed approach is evaluated on two different data sets. The obtained results demonstrate that it is suitable for modelling and monitoring multi-view streaming phenomena by providing means for continuous analysis and pattern mining. © 2021, IFIP International Federation for Information Processing.

The amount of data generated is on rise due to increased demand for fields like IoT, smart monitoring applications, etc. Data generated through such systems have many distinct characteristics like continuous data generation, evolutionary, multi-source nature, and heterogeneity. In addition, the real-world data generated in these fields is largely unlabelled. Clustering is an unsupervised learning technique used to group, analyze and interpret unlabelled data. Conventional clustering algorithms are not suitable for dealing with data having previously mentioned characteristics due to memory and computational constraints, their inability to handle concept drift, distributed location of data. Therefore novel clustering approaches capable of analyzing and interpreting evolving and/or multi-source streaming data are needed. 

The thesis is focused on building evolutionary clustering algorithms for data that evolves over time. We have initially proposed an evolutionary clustering approach, entitled Split-Merge Clustering (Paper I), capable of continuously updating the generated clustering solution in the presence of new data. Through the progression of the work, new challenges have been studied and addressed. Namely, the Split-Merge Clustering algorithm has been enhanced in Paper II with new capabilities to deal with the challenges of multi-view data applications. A multi-view or multi-source data presents the studied phenomenon/system from different perspectives (views), and can reveal interesting knowledge that is not visible when only one view is considered and analyzed. This has motivated us to continue in this direction by designing two other novel multi-view data stream clustering algorithms. The algorithm proposed in Paper III improves the performance and interpretability of the algorithm proposed in Paper II. Paper IV introduces a minimum spanning tree based multi-view clustering algorithm capable of transferring knowledge between consecutive data chunks, and it is also enriched with a post-clustering pattern-labeling procedure. 

The proposed and studied evolutionary clustering algorithms are evaluated on various data sets. The obtained results have demonstrated the robustness of the algorithms for modeling, analyzing, and mining evolving data streams. They are able to adequately adapt single and multi-view clustering models by continuously integrating newly arriving data. 

In smart buildings, many different systems work in coordination to accomplish their tasks. In this process, the sensors associated with these systems collect large amounts of data generated in a streaming fashion, which is prone to concept drift. Such data are heterogeneous due to the wide range of sensors collecting information about different characteristics of the monitored systems. All these make the monitoring task very challenging. Traditional clustering algorithms are not well equipped to address the mentioned challenges. In this work, we study the use of MV Multi-Instance Clustering algorithm for multi-view analysis and mining of smart building systems’ sensor data. It is demonstrated how this algorithm can be used to perform contextual as well as integrated analysis of the systems. Various scenarios in which the algorithm can be used to analyze the data generated by the systems of a smart building are examined and discussed in this study. In addition, it is also shown how the extracted knowledge can be visualized to detect trends in the systems’ behavior and how it can aid domain experts in the systems’ maintenance. In the experiments conducted, the proposed approach was able to successfully detect the deviating behaviors known to have previously occurred and was also able to identify some new deviations during the monitored period. Based on the results obtained from the experiments, it can be concluded that the proposed algorithm has the ability to be used for monitoring, analysis, and detecting deviating behaviors of the systems in a smart building domain. © 2021 by the authors. Licensee MDPI, Basel, Switzerland.