Time series analysis and prediction is a difficult and complex problem. Many machine- and deep-learning methods exist with better and better results. This paper proposes a strategy called Multi Model Recursion. It uses separate deep-learning models per feature that needs predicting. Another improvement is not predicting features which are easily calculated. Having extra models per feature helps in "simulating" a future environment since it predicts external variables otherwise unknown. The Multi Model Recursion developed is an improvement of the commonly used Recursive strategy. The paper compares this method with models and strategies frequently used in the field. The testing dataset is put together from publicly available Hungarian electricity load and weather data. The task was to predict the country's net electricity load for the next 3 hours.

Erlang tooling offers rich options to control the exact tasks to perform during an upgrade. This control aims to allow for zero-downtime upgrades. Upgrades affecting multiple dependent modules must reflect the dependency order in the upgrade's configuration, as an erroneous configuration results in unintended behaviour, possibly even downtime. This paper presents two static analysis-based checkers for verifying module-related aspects of upgrades. In our first analysis, we compare the actual dependency order derived from the source code with that expressed in the upgrade's configuration. We also analyze the code to find circular dependencies among its modules. These pose a problem during upgrades and are generally good practices to avoid.

Ensuring thread safety in applications is crucial for preventing subtle and challenging bugs in concurrent programming. This paper presents two algorithmic approaches to improve thread safety through static analysis and to demonstrate their benefits in real life, the authors also implemented them as two detectors in SpotBugs static analyzer. These checkers are designed to identify unsafe usages of shared resources and improper atomic operations in concurrent Java programming, aiming to mitigate common multithreading issues such as race conditions. By emphasizing consistent locking strategies and the correct use of atomic types, the study offers insight into how to improve the reliability of multithreaded applications.

Rapid changes, dynamic consumer preferences, and evolving market trends are the hallmarks of the business environment. SAP HANA has emerged as a potent platform to meet this demand due to its resilient foundation for real-time data analytics and processing and in-memory processing architecture. This research aims to improve anomaly detection capabilities by integrating machine learning (ML) models into the SAP HANA Fiori web application. This will be achieved by developing a custom Application Programming Interface (API). The proposed solution integrates ML models with SAP systems using FastAPI, providing real-time insights and decision-making capabilities, by leveraging scikit-learn's Local Outlier Factor (LOF) for anomaly detection. To guarantee seamless performance and scalability, the API is deployed on Azure using Docker containers. This paper illustrates the capability of custom APIs to integrate ML models into enterprise systems, enhance operational efficiency, and establish a reliable framework for real-time anomaly detection. The article addresses challenges associated with API integration, scalability, and system configuration, providing valuable insights for enhancing the deployment of machine learning in enterprise applications.

Due to their decentralized and trustless nature, blockchain and distributed ledger technologies are increasingly used in several domains, including critical applications. The behavior of such blockchain-integrated systems is typically driven by smart contracts. However, smart contracts are application-specific software and may contain faults with severe system-level impacts. This is especially true in the case of the extensively used Hyperledger Fabric (HLF) platform, where smart contracts are written in general-purpose languages (Java, among others), and applications can go far beyond handling virtual-currency-like assets. In this work, we present a novel formal-verification-based approach to smart contract verification and a high-level empirical model of the HLF platform. Our Smart Contract in the Loop (SCIL) method uses a model checker (Java Pathfinder) to check whether specific error properties hold for a given smart contract, while a predefined combination of platform-level fault modes is active. We facilitate the checking of HLF smart contracts without modification and enable the propagation or non-propagation of platform faults through the smart contracts to the system failure level.

The identification of the sources of a runtime error is a common task for Erlang developers. Dynamic and static tools can assist in this task. Our work aims to help Erlang developers in debugging processes to reproduce a runtime error. We would like to use and extend the static analyser framework of RefactorErl with new algorithms to support this fault localisation process. In our previous paper, we presented a symbolic execution-based analysis method to find the source of runtime errors. This paper extends that work with path selection heuristics to improve the efficiency of the algorithm in the RefactorErl framework.

Radiomics is an emerging field of CT image processing, that offers noninvasive quantification of tumour phenotypes using quantitative image features. Radiomics analysis has promising applications in cancer treatment and personalized medicine, like treatment planning and the prediction of clinical factors. However, the optimal feature selection is not established in the literature, and the applications usually involve data mining of a large pool of features. In this paper, we propose to extract higher-level radiomic features using Radial Harmonic Fourier moments (RHFM). Image moments, and specially orthogonal Fourier moments are widely used in image processing, providing efficient and invariant shape descriptors. In particular, RHFMs are known to perform well on small noisy images, making them a promising candidate for CT tumour analysis. Motivated by these advantages, we developed a feature extraction scheme based on RHFM, and we performed radiomics analysis on lung CT images of non-small cell lung cancer patients. The proposed method is validated on multiple annotated datasets following the literature guidelines, evaluating the accuracy, stability, reliability, and prognostic value of the proposed features. The results show better reliability and otherwise comparable performance compared to the state-of-the-art wavelet descriptors. Furthermore, Fourier moments provide higher level of flexibility and possible adaptivity compared to wavelets, and unlike wavelet features, RHFM features are invariant of position, size and orientation in the tumor region.

Due to the lockdowns caused by the COVID-19 pandemic, the majority of educational institutions worldwide have been forced to switch to online education, which has created a significant challenge for teachers and students alike. In order to communicate effectively in the online space, educational institutions had a wide range of tools to choose from (e.g. Adobe Connect, Cisco Webex, Google Meet, Microsoft Teams, Skype, Zoom, etc.). The challenge for teachers was to learn how to use them, to teach practical subjects effectively and to provide a supervised examination environment. The return to face-to-face (in-class) teaching after the end of the COVID-19 pandemic has allowed the online collaborative environments listed above to fade into the background, but the supervision of interactive, computer-based practical lessons (e.g. teaching programming languages, network programming etc.) and proctored examinations can still be a challenge for teachers. This article reviews some screen monitoring systems developed for both corporate and educational environments. We present one of them in more detail, namely Veyon, which is available free of charge\footnote{The basic version of Veyon is free, but you have to pay for the various desired add-on licenses.} and can be used on different operating systems, and whose applicability in both teaching and examination has been tested for almost a year at the Faculty of Informatics of the University of Debrecen.

Due to the importance of depth perception in virtual spaces, the combined effects of display devices and human factors on egocentric distance estimation were investigated. We developed a virtual environment that can assess distance estimation skills of users at 10 various distances, starting from 25 cm and ending at 160 cm. Our results show that people are either accurate or overestimate distances on a desktop display, while underestimation occurs with the Gear VR in most cases. Combined with display devices, human factors also had effects on distance estimates. With the Gear VR, 35.73%-57.14% faster estimation times were obtained, and these can also be influenced by human factors and distances.

In our paper we present a parallel co-reference analysis of the Hungarian poem A szőlőműves (The Vine-Dresser) by Milán Füst and its English translation. We explore the textual world of the poem and compare the Hungarian and English texts using formal linguistic tools based on semiotic textology. We also reveal the possible differences between the original Hungarian text and its English translation. The results of the analysis prove that co-reference analysis devised and elaborated by János S. Petőfi can be effectively applied in a polyglot environment. We also introduce the hypertext implementation of the co-reference analy-sis of the selected poetic texts in a form of a web page. We would like to show that this has many practical advantages, for example, the analysis and its results can be transparent, accessible, and verifiable to everyone. In addition, the created web page provides additional aspects for the analysis. For example, considering the poetic text as a network, we can investigate whether the scale-free feature is also relevant in the textological environment – as it can be experienced in many other areas of reality.