A growing number of research papers shed light on automated machine learning (AutoML) frameworks, which are becoming a promising solution for building complex machine learning models without human expertise and assistance. The key challenge in enabling AutoML frameworks to build an efficient model for anomaly detection tasks is to determine the best underlying model for a given task and optimization metric. The meta-learning approaches based on a set of meta-features that describes data properties can enable efficient model selection in AutoML frameworks. The existing meta-learning approaches based on statistical and information-theoretic meta-features require large amounts of data and computational resources to extract data properties. This paper proposes a novel set of meta-features for model selection in anomaly detection tasks based on domain-specific properties of data which overcomes the shortcomings of existing meta-features by introducing simple but effective meta-features that can be efficiently extracted or estimated by using a low amount of data. Experiments with 63 datasets from different repositories with varying schemas show that the proposed set of meta-features achieves an accuracy of 87% for model selection, while the achieved accuracy for simple meta-features is 74%, for statistical meta-features 68%, for information theory meta-feature 70%, and for a comprehensive set of meta-features by pyMFE 73%. This demonstrates that the proposed set can be adopted by AutoML frameworks across a diverse range of domains.

Economic development, as one of the three pillars of sustainability, can be enhanced by utilizing open-source software. The impact of the pandemic on software development and whether or not it managed to sustain the velocity and volume it previously had has already piqued the interest of the research community. From measuring the activity of developers to conducting surveys on the perceived productivity, the research was mostly focused on 2020. This paper focuses on how the pandemic impacted public development on GitHub by analyzing the changes in trends during the 2020–2022 period (COVID-19 period) compared with 2017–2019 (pre-COVID-19 period). While the majority of events have continued with relatively the same trend, having minor increases or decreases, there have been a few that stood out. Several events related to the community activity on GitHub experienced decreases in their trends (ForkEvent’s trend decreased $0.09$×, IssuesEvent $0.01$×, etc.), while events related to discussions have experienced a positive trend (mainly CommitCommentEvent, with a trend increase of 34×, but also IssueCommentEvent, which remained stable despite the $0.01$× decrease of IssuesEvent), but only CommitCommentEvent and PushEvent (which experienced a $2.52$× increase in its trend) exhibited non-stationary behavior in the ADF test. In general, events related to individual development have sustained or increased their trends, while events related to community activity (i.e., forking) or contributions to documentation have somewhat decreased. We believe this shows that although there have been minor reductions in the casual and community contributions on GitHub, the majority of events continued with the same trend or even with some increases, which shows that GitHub developers persevered in the face of the pandemic.

Parkinson’s disease (PD) is a progressive neurodegenerative disorder in which early detection remains a major clinical challenge due to heterogeneous motor and non-motor manifestations and the lack of reliable biomarkers. In recent years, machine learning (ML) and deep learning (DL) methods have been increasingly investigated as decision-support tools for PD screening using diverse clinical and behavioral data. This review synthesizes PD detection studies published between 2017 and 2025, systematically analyzing 32 representative works across multiple modalities, including MRI, PET, EEG, REM sleep biomarkers, voice recordings, gait signals, handwriting/drawing tasks, and finger-tapping measurements. Across the reviewed literature, high classification performance is frequently reported, with CNN-based and hybrid DL architectures achieving particularly strong results in imaging and time-series settings, while classical ML approaches such as SVM and ensemble models remain competitive for engineered feature-based datasets. However, the review also reveals major barriers to reliable translation, including small datasets, inconsistent evaluation protocols, limited external validation, and the risk of performance inflation caused by non-subject-independent data splitting. Overall, this review provides a structured and modality-oriented reference of algorithms, datasets, and performance trends, while highlighting key methodological gaps and practical priorities for developing robust and clinically deployable PD detection systems.

The selection of code reviewers for a pull request can impact the quality as well as the speed of the review. In order to have the best experience both for the developer and the reviewers, there is a need for automatic reviewer recommendation systems for pull requests. Although there has been previous research in this area, it was mainly focused on smaller and medium repositories of up to around 200 developers, while larger repositories were rarely targeted. In this paper we evaluate several existing approaches on a set of 8 Microsoft repositories of different sizes, noticing that the average performance of the approaches seemed to decrease with the number of reviewers the repository has. In order to focus the existing approaches only on relevant reviewers, we propose a technique for improving their performance by scoping down the set of candidate reviewers based on multiple filters. We defined several basic filters and determined that 5 out of the 7 tested existing approaches experienced performance improvements of up to 16.24% better precision and 19.66% better recall averaged over all datasets, with the per dataset improvements peaking at 36.63% better precision and 28.63% better recall. Additionally, by combining different basic filters we were able to achieve additional improvements for 6 out of the 7 existing approaches (with a minor improvement for the remaining approach), which on average over all datasets had up to 17.60% better precision and 21.23% better recall, while the per dataset improvements peaked at 43.28% better precision and 30.94% better recall.

Describing architectures of complex software systems using architectural languages is usually done through multiple viewpoints that enable the creation of views. While the creation of views enables the separation of stakeholders’ concerns with the system and eases manageability, it raises the problem of inconsistencies among the views. This paper presents Annotated Functional Decomposition (AFD), an architectural language that provides integral modeling as a possible solution to this problem. Integral modeling creates a model by decomposing a system into its functions, which are annotated to simultaneously create multiple views. Having all created views available in the model at the same time facilitates inconsistency management. AFD supports automated inconsistency detection and manual inconsistency resolution. Moreover, AFD supports the automated translation of views to appropriate UML diagrams, which facilitates adaptation to other methodological approaches. According to the criteria used in the literature for the evaluation of 124 architectural languages, AFD provides nine out of 12 requirements that are important to practitioners.