Predictive modeling has the potential to improve risk stratification of cancer patients and thereby contribute to optimized treatment strategies and better outcomes for patients in clinical practice. To develop robust predictive models for decision-making in healthcare, sensitive patient-level data is often required when developing the training models. Consequently, data privacy is an important aspect to consider when building these predictive models and in subsequent communication of the results. In this study we have used Graph Neural Networks for survival prediction, and compared the accuracy to state-of-the-art prediction models after applying Differential Privacy and k-Anonymity, i.e. two privacy-preservation solutions. By using two different data sources we demonstrated that Graph Neural Networks and Survival Forests are the two most well-performing survival prediction methods when used in combination with privacy preservation solutions. Furthermore, when the predictive model was built using clinical expertise in the specific area of interest, the prediction accuracy of the proposed knowledge based graph model drops by at most 10% when used with privacy preservation solutions. Our proposed knowledge based graph is therefore more suitable to be used in combination with privacy preservation solutions as compared to other graph models.

Traffic congestion is a problem in day to day life, especially in big cities. Various traffic control infrastructure systems have been deployed to monitor and improve the flow of traffic across cities. Real-time congestion detection can serve for many useful purposes that include sending warnings to drivers approaching the congested area and daily route planning. Most of the existing congestion detection solutions combine historical data with continuous sensor readings and rely on data collected from multiple sensors deployed on the road, measuring the speed of vehicles. While in our work we present a framework that works in a pure streaming setting where historic data is not available before processing. The traffic data streams, possibly unbounded, arrive in real-time. Moreover, the data used in our case is collected only from sensors placed on the intersections of the road. Therefore, we investigate in creating a real-time congestion detection and reduction solution, that works on traffic streams without any prior knowledge. The goal of our work is 1) to detect traffic jams in real-time, and 2) to reduce the congestion in the traffic jam areas.In this work, we present a real-time traffic jam detection and congestion reduction framework: 1) We propose a directed weighted graph representation of the traffic infrastructure network for capturing dependencies between sensor data to measure traffic congestion; 2) We present online traffic jam detection and congestion reduction techniques built on a modern stream processing system, i.e., Apache Flink; 3) We develop dynamic traffic light policies for controlling traffic in congested areas to reduce the travel time of vehicles. Our experimental results indicate that we are able to detect traffic jams in real-time and deploy new traffic light policies which result in 27% less travel time at the best and 8% less travel time on average compared to the travel time with default traffic light policies. Our scalability results show that our system is able to handle high-intensity streaming data with high throughput and low latency.

Privacy preservation plays a vital role in health care applications as the requirements for privacy preservation are very strict in this domain. With the rapid increase in the amount, quality and detail of health data being gathered with smart devices, new mechanisms are required that can cope with the challenges of large scale and real-time processing requirements. Federated learning (FL) is one of the conventional approaches that facilitate the training of AI models without access to the raw data. However, recent studies have shown that FL alone does not guarantee sufficient privacy. Differential privacy (DP) is a well-known approach for privacy guarantees, however, because of the noise addition, DP needs to make a trade-off between privacy and accuracy. In this work, we design and implement an end-to-end pipeline using DP and FL for the first time in the context of health data streams. We propose a clustering mechanism to leverage the similarities between users to improve the prediction accuracy as well as significantly reduce the model training time. Depending on the dataset and features, our predictions are no more than 0.025% far off the ground-truth value with respect to the range of value. Moreover, our clustering mechanism brings a significant reduction in the training time, with up to 49% reduction in prediction accuracy error in the best case, as compared to training a single model on the entire dataset. Our proposed privacy preserving mechanism at best introduces a decrease of ≈ 2% in the prediction accuracy of the trained models. Furthermore, our proposed clustering mechanism reduces the prediction error even in highly noisy settings by as much as 38% as compared to using a single federated private model.

Neural networks are used for different machine learning tasks, such as spatial time-series forecasting. Accurate modelling of a large and complex system requires large datasets to train a deep neural network that causes a challenge of scale as training the network and serving the model are computationally and memory intensive. One example of a complex system that produces a large number of spatial time-series is a large road sensor infrastructure deployed for traffic monitoring. The goal of this work is twofold: 1) To model large amount of spatial time-series from road sensors; 2) To address the scalability problem in a real-life task of large-scale road traffic prediction which is an important part of an Intelligent Transportation System.We propose a partitioning technique to tackle the scalability problem that enables parallelism in both training and prediction: 1) We represent the sensor system as a directed weighted graph based on the road structure, which reflects dependencies between sensor readings, and weighted by sensor readings and inter-sensor distances; 2) We propose an algorithm to automatically partition the graph taking into account dependencies between spatial time-series from sensors; 3) We use the generated sensor graph partitions to train a prediction model per partition. Our experimental results on traffic density prediction using Long Short-Term Memory (LSTM) Neural Networks show that the partitioning-based models take 2x, if run sequentially, and 12x, if run in parallel, less training time, and 20x less prediction time compared to the unpartitioned model of the entire road infrastructure. The partitioning-based models take 100x less total sequential training time compared to single sensor models, i.e., one model per sensor. Furthermore, the partitioning-based models have 2x less prediction error (RMSE) compared to both the single sensor models and the entire road model.

Graph partitioning is an essential yet challenging task for massive graph analysis in distributed computing. Common graph partitioning methods scan the complete graph to obtain structural characteristics offline, before partitioning. However, the emerging need for low-latency, continuous graph analysis led to the development of online partitioning methods. Online methods ingest edges or vertices as a stream, making partitioning decisions on the fly based on partial knowledge of the graph. Prior studies have compared offline graph partitioning techniques across different systems. Yet, little effort has been put into investigating the characteristics of online graph partitioning strategies. In this work, we describe and categorize online graph partitioning techniques based on their assumptions, objectives and costs. Furthermore, we employ an experimental comparison across different applications and datasets, using a unified distributed runtime based on Apache Flink. Our experimental results showcase that model-dependent online partitioning techniques such as low-cut algorithms offer better performance for communication-intensive applications such as bulk synchronous iterative algorithms, albeit higher partitioning costs. Otherwise, model-agnostic techniques trade off data locality for lower partitioning costs and balanced workloads which is beneficial when executing data-parallel single-pass graph algorithms.