Smartphones allow for prompting users with a short questionnaire about their current subjective experience, a technique often called Ecological Momentary Assessments. One of the biggest challenges for such studies is a lack of adherence, diminishing the benefits for both user and researcher. Being able to predict if a user is going to stop answering the questionnaire prompts would be beneficial for researchers and developers. This would allow for, for example, specifically addressing those users, or for over-sampling populations at higher risk of dropping out of a study. In this work, based on an observational study of the general population, we analyzed data from almost 1,000 users. The data include a large variety of sensor data from the users' smartphones. We utilized machine learning to predict adherence on a day-to-day level, as well as predict adherence based on participant data after on-boarding. For day-to-day prediction, the best performing model was a model based on metadata features (days since first questionnaire was filled out, days since the last questionnaire was filled out, number of filled-out questionnaires, days since app installation), yielding an area under the precision-recall curve of 0.89. The inclusion of sensor data did not improve the model's performance, indicating that the high cost of collecting and processing sensor data is not worth the benefits for predicting fill-out behavior. Predicting at sign-up if a user will adhere to a questionnaire prompt at least once was better than chance, but further studies are needed.

The challenge of sustaining user engagement in eHealth interventions is a pressing issue with significant implications for the effectiveness of these digital health tools. This study investigates user engagement in a cognitive-behavioral therapy-based eHealth intervention for procrastination, using a dataset from a randomized controlled trial of 233 university students. Various machine learning models, including Decision Tree, Gradient Boosting, Logistic Regression, Random Forest, and Support Vector Machines, were employed to predict patterns of user engagement. The study adopted a two-phase analytical approach. In the first phase, all features of the dataset were included, revealing ‘total_minutes’—the total time participants spent on the intervention and the eHealth platform—as the most significant predictor of engagement. This finding emphasizes the intuitive notion that early time spent on the platform and the intervention is a strong indicator of later user engagement. However, to gain a deeper understanding of engagement beyond this predominant metric, the second phase of the analysis excluded ‘total_minutes’. This approach allowed for the exploration of the roles and interdependencies of other engagement indicators, such as ‘number_intervention_answersheets’—the number of completed lessons, ‘logins_first_4_weeks’—login frequency, and ‘number_diary_answersheets’—the number of completed diaries. The results from this phase highlighted the multifaceted nature of engagement, showing that while ‘total_minutes’ is strongly correlated with engagement, indicating that more engaged participants tend to spend more time on the intervention, the comprehensive engagement profile also depends on additional aspects like lesson completions and frequency of platform interactions.

Notifications are an essential part of the user experience on smart mobile devices. While some apps have to notify users immediately after an event occurs, others can schedule notifications strategically to notify them only on opportune moments. This tailoring allows apps to shorten the users’ interaction delay. In this paper, we present the results of a comprehensive study that identified the factors that influence users’ interaction delay to their smartphone notifications. We analyzed almost 10 million notifications collected in-the-wild from 922 users and computed their response times with regard to their demographics, their Big Five personality trait scores and the device’s charging state. Depending on the app category, the following tendencies can be identified over the course of the day: Most notifications were logged in late morning and late afternoon. This number decreases in the evening, between 8 p.m. and 11 p.m., and at the same time exhibits the lowest average interaction delays at daytime. We also found that the user’s sex and age is significantly associated with the response time. Based on the results of our study, we encourage developers to incorporate more information on the user and the executing device in their notification strategy to notify users more effectively.

Summarization for scientific text has shown significant benefits both for the research community and human society. Given the fact that the nature of scientific text is distinctive and the input of the multi-document summarization task is substantially long, the task requires sufficient embedding generation and text truncation without losing important information. To tackle these issues, in this paper, we propose SKT5SciSumm - a hybrid framework for multi-document scientific summarization (MDSS). We leverage the Sentence-Transformer version of Scientific Paper Embeddings using Citation-Informed Transformers (SPECTER) to encode and represent textual sentences, allowing for efficient extractive summarization using k-means clustering. We employ the T5 family of models to generate abstractive summaries using extracted sentences. SKT5SciSumm achieves state-of-the-art performance on the Multi-XScience dataset. Through extensive experiments and evaluation, we showcase the benefits of our model by using less complicated models to achieve remarkable results, thereby highlighting its potential in advancing the field of multi-document summarization for scientific text.

During the COVID-19 pandemic, the novel coronavirus had an impact not only on public health but also on the mental health of the population. Public sentiment on mental health and depression is often captured only in small, survey-based studies, while work based on Twitter data often only looks at the period during the pandemic and does not make comparisons with the pre-pandemic situation. We collected tweets that included the hashtags #MentalHealth and #Depression from before and during the pandemic (8.5 months each). We used LDA (Latent Dirichlet Allocation) for topic modeling and LIWC, VADER, and NRC for sentiment analysis. We used three machine-learning classifiers to seek evidence regarding an automatically detectable change in tweets before vs. during the pandemic: (1) based on TF-IDF values, (2) based on the values from the sentiment libraries, (3) based on tweet content (deep-learning BERT classifier). Topic modeling revealed that Twitter users who explicitly used the hashtags #Depression and especially #MentalHealth did so to raise awareness. We observed an overall positive sentiment, and in tough times such as during the COVID-19 pandemic, tweets with #MentalHealth were often associated with gratitude. Among the three classification approaches, the BERT classifier showed the best performance, with an accuracy of 81% for #MentalHealth and 79% for #Depression. Although the data may have come from users familiar with mental health, these findings can help gauge public sentiment on the topic. The combination of (1) sentiment analysis, (2) topic modeling, and (3) tweet classification with machine learning proved useful in gaining comprehensive insight into public sentiment and could be applied to other data sources and topics.

At the beginning of the COVID-19 pandemic, with a lack of knowledge about the novel virus and a lack of widely available tests, getting first feedback about being infected was not easy. To support all citizens in this respect, we developed the mobile health app Corona Check. Based on a self-reported questionnaire about symptoms and contact history, users get first feedback about a possible corona infection and advice on what to do. We developed Corona Check based on our existing software framework and released the app on Google Play and the Apple App Store on April 4, 2020. Until October 30, 2021, we collected 51,323 assessments from 35,118 users with explicit agreement of the users that their anonymized data may be used for research purposes. For 70.6% of the assessments, the users additionally shared their coarse geolocation with us. To the best of our knowledge, we are the first to report about such a large-scale study in this context of COVID-19 mHealth systems. Although users from some countries reported more symptoms on average than users from other countries, we did not find any statistically significant differences between symptom distributions (regarding country, age, and sex). Overall, the Corona Check app provided easily accessible information on corona symptoms and showed the potential to help overburdened corona telephone hotlines, especially during the beginning of the pandemic. Corona Check thus was able to support fighting the spread of the novel coronavirus. mHealth apps further prove to be valuable tools for longitudinal health data collection.

Stress is leading to bad health and contributes to economic loss due to employee absence. Real-time stress detection based on wearable sensor data can enable the implementation of mitigating strategies. While several approaches to stress detection exist, setting up a new system can be tedious. We demonstrate how the use of libraries and tools for automation can speed up many of the necessary steps when developing a stress detection system. We employ automated feature engineering and automated machine learning. The resulting stress detection system we developed this way is based on the WESAD dataset and achieves a F1 score of 0.87 for unseen users based on 30 seconds of wearable sensor data.

Risk prediction in patients with heart failure (HF) is essential to improve the tailoring of preventive, diagnostic, and therapeutic strategies for the individual patient, and effectively use health care resources. Risk scores derived from controlled clinical studies can be used to calculate the risk of mortality and HF hospitalizations. However, these scores are poorly implemented into routine care, predominantly because their calculation requires considerable efforts in practice and necessary data often are not available in an interoperable format. In this work, we demonstrate the feasibility of a multi-site solution to derive and calculate two exemplary HF scores from clinical routine data (MAGGIC score with six continuous and eight categorical variables; Barcelona Bio-HF score with five continuous and six categorical variables). Within HiGHmed, a German Medical Informatics Initiative consortium, we implemented an interoperable solution, collecting a harmonized HF-phenotypic core data set (CDS) within the openEHR framework. Our approach minimizes the need for manual data entry by automatically retrieving data from primary systems. We show, across five participating medical centers, that the implemented structures to execute dedicated data queries, followed by harmonized data processing and score calculation, work well in practice. In summary, we demonstrated the feasibility of clinical routine data usage across multiple partner sites to compute HF risk scores. This solution can be extended to a large spectrum of applications in clinical care.

The application of machine learning algorithms has become important for the medical domain. However, the concrete application of these type of algorithms strongly depends on how a corresponding data source was created. Most importantly, domain knowledge must be linked with data science knowledge. Data collected using smartphones or smart mobile devices (e.g., smart watches) is commonly referred to as mHealth data. The possibilities and strategies for collecting data in this area now appear to be as diverse as the machine learning algorithms that have emerged. This tutorial will therefore discuss how mHealth data is structured and which aspects need to be taken into account when evaluating it with machine learning algorithms, using concrete examples.

The microservice architectural style has gained much attention from both academia and industry recently as a novel way to design, develop, and deploy cloud-native applications. This concept encourages the decomposition of a monolith into multiple independently deployable units. A typical microservices-based application is formed of two service types: functional services, which provide the core business logic, and infrastructure services, which provide essential functionalities for a microservices ecosystem. To improve developers' productivity, many software frameworks have been developed to provide those reusable infrastructure services, allowing programmers to focus on implementing microservices in arbitrary ways. In this work, we made use of four open source frameworks to develop a cloud-based application in order to compare and evaluate their usability and practicability. While all selected frameworks promote asynchronous microservice design in general, there are differences in the ways each implements services. This leads to interoperability issues, such as message topic naming convention. Additionally, a key finding is the long startup times of JVM-based services that might reduce application's resiliency and portability. Some other advantages come directly from the programming language, such as the ability of Go to generate native binary executables, which results in very small and compact Docker images (up to 78% smaller compared to other languages).

In the last few years, the proliferation of edge and cloud computing infrastructures as well as the increasing number of mobile devices has facilitated the emergence of many novel applications. However, that increase of complexities also creates novel challenges for service providers, for example, the efficient management of interdependent services during runtime. One strategy is to reallocate services dynamically by migrating them to suitable servers. However, not every microservice can be deployed as stateless instances, which leads to suboptimal performance of live migration techniques. In this work, we propose a novel live migration scheme focusing on stateful microservices in edge/cloud environments by utilizing the underlying messaging infrastructure to reconstruct the service’s state. Not only can this approach be applied in various microservice deployment scenarios, experimental evaluation results also show a reduction of 19.92% downtime compared to the stop-and-copy migration method.

With about 300 million affected people, major depressive disorder (MDD) is one of the most common diseases worldwide. During the COVID-19 pandemic, the number of cases increased even further, by 28%. Many factors may be correlated with MDD, including the excessive use of social media apps. In this paper, we investigated the relationship between the use of social media and communication apps and depressive symptoms during the COVID-19 pandemic. The pandemic and social distancing like lockdowns probably changed smartphone usage times and usage patterns. While previous studies have shown an association between depression and social media usage, we report about the situation during these special circumstances. We employed a log-linear regression to examine the association of social media and communication app usage and depression. To quantify the usage, we applied the total usage time in hours of social media apps (e.g., WhatsApp, Facebook) as well as communication apps (Phone and Messaging) within one week. To measure depressive symptoms, we used the PHQ-9 score. We discovered a significant association between the usage time and the PHQ-9 score (beta=0.0084, p-value=0.010). We conclude that social media usage is a robust marker for depression severity and future research should focus on a better understanding of the underlying causality and potential counter-measures.

Next app prediction can help enhance user interface design, pre-loading of apps, and network optimizations. Prior work has explored this topic, utilizing multiple different approaches but challenges like the user cold-start problem, data sparsity, and privacy concerns related to contextual data like location histories, persist. The user cold-start problem occurs when a user has recently registered to the smartphone app system and there is not enough information about his/her preferences and his/her history of smartphone usage. In this work, we try to address the above issues. We introduce WhatsNextApp, an approach based on LSTM (Long Short-Term Memory) networks using sequences of app usage logs. Our approach is inspired by Word Embeddings and treats sequences of app usage logs as sequences of words. We collect a real-life data set consisting of 975 Android users with over 22 million app usage events. We build a generic (user-independent) WhatsNextApp model and the evaluation with our data set shows that it outperforms related studies for existing users where we achieve a recall@8 (recall for the top 8 apps) of 92%. For the user cold-start problem with the 500 most frequent apps, we achieve a recall@8 of 82.7%.

This chapter gives a brief summary about the first part of the thesis and transitions to the second part of the thesis.

This chapter introduces TYDR—Track Your Daily Routine, an Android app that enables the research of the relationship between smartphone sensor data and usage statistics and the users' personality via mobile crowdsensing. TYDR processes and visualizes the tracked data as well as the results of the personality questionnaires. We released TYDR on Google Play in October 2018 and registered 3,921 installations since then. To the best of our knowledge, we are the first to propose a full-scale, integrated privacy model specifically for apps related to mobile collection of data. Our privacy model called PM-MoDaC (Privacy Model for Mobile Data Collection Apps) consists of nine concrete measures to be taken to ensure the users' privacy. These measures include transparently informing the users about the data being collected, the anonymization of user data, and enabling an opt-out option. We present the implementation of all the privacy measures in TYDR. We collected data with TYDR over the course of a two-months period and extensively evaluated our privacy model and which data users are willing to share. Based on data about granting TYDR all or no Android system permissions, we found evidence that younger users tend to be less willing to share their data (average age of 30 years compared to 35 years). We also observed that female users tend to be less willing to share data compared to male users. We did not find any evidence that education or personality traits are a factor related to data sharing. TYDR users score higher on the personality trait openness to experience than the average of the population, which we assume to be evidence that the type of app influences the user base it attracts in terms of average personality traits.

This chapter gives an extensive overview about related work in the field of ubiquitous social networking. We start with presenting the fundamentals about related concepts from psychology and social sciences. The key finding is that similarity is the key feature structuring social networks. Then, the chapter gives details about relevant papers of the last 15 years, highlighting the research questions addressed and topics covered. One of the most common use case scenarios is the incentivization of social interaction. On a technical level, common themes are software architectures and device-to-device communication.

For a non-technical audience, new concepts from computer science and engineering are often hard to grasp. In order to introduce a general audience to topics related to Industry 4.0, we designed and developed a demonstrator game. The Who-wants-to-be-a-millionaire?-style quiz game lets the player experience indoor positioning based on Bluetooth signal strength firsthand. We found that such an interactive game demonstrator can function as a conversation-opener and is useful in helping introduce concepts relevant for many future jobs.

Through lockdowns and other severe changes to daily life, almost everyone is affected by the COVID-19 pandemic. Scientists and medical doctors are - among others - mainly interested in researching, monitoring, and improving physical and mental health of the general population. Mobile health apps (mHealth), and apps conducting ecological momentary assessments (EMA) respectively, can help in this context. However, developing such mobile applications poses many challenges like costly software development efforts, strict privacy rules, compliance with ethical guidelines, local laws, and regulations. In this paper, we present TrackYourHealth (TYH), a highly configurable, generic, and modular mobile data collection and EMA platform, which enabled us to develop and release two mobile multiplatform applications related to COVID-19 in just a few weeks. We present TYH and highlight specific challenges researchers and developers of similar apps may also face, especially when developing apps related to the medical field.