To achieve the full analytical potential of the streaming data from the internet of things, the interconnection of various data sources is needed. By definition, those sources are heterogeneous and their integration is not a trivial task. A common approach to exploit streaming sensor data potential is to use machine learning techniques for predictive analytics in a way that is agnostic to the domain knowledge. Such an approach can be easily integrated in various use cases. In this paper, we propose a novel framework for data fusion of a set of heterogeneous data streams. The proposed framework enriches streaming sensor data with the contextual and historical information relevant for describing the underlying processes. The final result of the framework is a feature vector, ready to be used in a machine learning algorithm. The framework has been applied to a cloud and to an edge device. In the latter case, incremental learning capabilities have been demonstrated. The reported results illustrate a significant improvement of data-driven models, applied to sensor streams. Beside higher accuracy of the models the platform offers easy setup and thus fast prototyping capabilities in real-world applications.

Using synthetic data sets to train medicine-focused machine learning models has been shown to enhance their performance; however, most research focuses on English texts. In this paper, we explore generating non-English synthetic medical texts. We propose a methodology for generating medical synthetic data, showcasing it by generating medical texts written in a non-English mixed language. We evaluate our approach with thirteen different language models that are open-source and proprietary, and assess the quality of the data sets in two ways: performing a statistical comparison between the original data set and the generated data sets, and training a classifier to distinguish between original and synthetic examples. The Llama-3.2-3B model achieves the best F1 score of 0.821 ± 0.007 and accuracy of 0.816 ± 0.016, making it most suitable for generating indistinguishable medical synthetic data. In contrast, models like Aya-23, Phi-3, and SmoLLM variants achieve high F1 scores (0.945–0.948), indicating their synthetic data is easily distinguishable from original data. These findings highlight the importance of model selection when generating synthetic medical data sets in non-English languages.

Efficient screening of transactions provides an empowering tool for anti-money laundering procedures and actions. Automatic classification and detection of anomalous behaviours and transaction structures enable faster and more effective action on the side of the supervisory authority. This chapter introduces research achievements and tools developed to streamline transaction monitoring and ease domain experts with automatic and semi-automatic filtering of risky transaction typologies. Presented tools are integrated as part of PAMLS (Platform for Anti-Money Laundering Supervision) to streamline and automate the discovery of risky behaviours in bank transaction data enriched with relevant company information. Enriched transactional data is pseudo-anonymized with respect to the legal and regulatory framework. Screening tool as a part of PAMLS platform automatically detects and marks specific predefined scenarios using newly developed state-of-the-art AI method tailored specifically to time-evolving transaction graphs in transaction data. Easy-to-use tools, early warning system and subsequent parameterized queries with additional white-listed scenarios provide domain experts with additional data to easily explore suggested dangerous transaction groups and make more informed decisions and further action, be it at a level of a specific financial institution or a cluster of them.

Modern text generation metrics use semantic representations of words to assess the quality of a text generation model without considering the fluency of the generated text. This paper proposes a novel text generation metric that combines adequacy and fluency to measure the quality of the generated text. When computing the final score using optimal transport, the metric considers semantic meaning and word order. We evaluate the metric on text translation data sets consisting of 20 language pairs from various language families and scripts. Using a novel statistic for measuring word order sensitivity, we analyze its adequacy-based performance using Pearson’s r and Kendall’s $\tau $ correlation coefficients and their sensitivity to fluency-related modifications. Results show that the proposed metric is the most sensitive to fluency-related changes among all top-performing embedding-based metrics, which were found to be relatively invariant to variations in word order. The proposed metric’s overall adequacy-based performance is lower than the best embedding-based metric but higher than the n-gram matching metrics. Our code is publicly available on GitHub ( https://github.com/eriknovak/metric-OPWScore ) under the BSD-2-Clause license.

Due to sensor failures and other issues, real-world time series may contain missing values, often in consecutive segments. Classification of such time series is an important task with prominent applications in various domains such as medicine, manufacturing, social networks and environmental sciences. In this paper, we consider various approaches that have been designed for this task, in particular, fully-convolutional neural networks (FCNs) with sparsity-invariant convolution and dynamic time warping convolution. We compare their performance to that of a standard transformer, TARNet, which has not been tailored to the classification of time series with missing values. Our results indicate that even this simple transformer may outperform the aforementioned models that were designed to deal with missing values. As this observation is consistent for many datasets from various domains and various distributions of missing values, we conclude that transformers are an exceptionally strong baseline for the classification of time series with missing values. In order to support the reproduction of our results as well as follow-up works, we performed the aforementioned experiments on publicly available time series datasets using a publicly available implementation of TARNet.