Anomaly detection and the identification of elements that do not fit the data characteristics are increasingly used in information systems, both for data cleaning and for finding unusual elements. Unsupervised anomaly detection methods are particularly useful in this context. This paper introduces the Optimized Deep Isolation Forest (ODIF) as an optimized version of the Deep Isolation Forest (DIF) algorithm. The training of DIF is subjected to an optimization of the operations performed, which leads to a reduction of the computational and memory complexity. In a series of experiments, both DIF and ODIF are implemented, and their effectiveness is evaluated using Area Under the Precision-Recall Curve (PR AUC). The proposed method demonstrates significantly better detection performance compared to the baseline Isolation Forest and competitive techniques. Additionally, the execution times of the training phase are measured for both the CPU and GPU stages, as well as memory usage, including RAM and VRAM. The results unequivocally indicate a much faster execution of the ODIF algorithm compared to DIF, with average CPU stage and GPU stage times being over one and a half times and nearly 150 times shorter, respectively. Similarly, memory usage is significantly reduced for ODIF in comparison to DIF, with RAM consumption lowered by approximately 18% and VRAM by over 55%. • We propose the Optimized Deep Isolation Forest (ODIF) method for anomaly detection. • ODIF achieves significant reduction in execution times for both CPU and GPU stages. • ODIF enhances memory efficiency with substantially reduced RAM and VRAM usage. • ODIF maintains detection effectiveness while optimizing computational complexity.

The COVID-19 pandemic triggered a sudden shift toward remote and hybrid work, placing IT technologies at the forefront of organizational practices and revealing a spectrum of emotional responses among employees. While conventional wisdom suggests that high salaries in the IT industry safeguard well-being, this study challenges the notion that “money can't buy happiness” by demonstrating that technostress, social isolation, and “Zoom fatigue” persist regardless of income level. Drawing on a longitudinal dataset collected at three one-year intervals, the research employs fuzzy semantics to translate qualitative survey data into quantitative descriptors, basket analysis to identify consistent behavioral patterns, and a Minimal Spanning Tree-Based Isolation Forest enhanced by Takagi–Sugeno fuzzy rules to detect anomalies. The findings indicate a nuanced interplay of negative and positive emotions, with fear, anxiety, and fatigue frequently coexisting alongside pride, energy, and satisfaction. Some anomalies in responses reveal issues such as random guessing or minimal IT usage, underscoring the importance of filtering out low-quality data. Crucially, the emotional outcomes are influenced not only by pandemic-related disruption but also by psychosocial and organizational factors – such as role complexity, skill requirements, and managerial support – providing a holistic view of how and why IT-based work can foster both strain and fulfillment. These insights hold practical implications for employers, suggesting that comprehensive well-being strategies, rather than monetary incentives alone, are pivotal in promoting a healthier, more resilient IT workforce.

Anomaly detection is one of the most important problems of modern data science due to the threat to the security of information systems as well as their users. This applies in particular to logistic data, which is used to predict costs, times, and organization of travel routes. Data anomalies may endanger the welfare and safety of transport users, goods, handling companies, and consumers. Moreover, they contribute to the overexploitation of the natural environment. Therefore, it is extremely important to find methods that are responsible for their effective detection. The desired approach may be the Choquet integral and its extensions, which in various applications have proven that with their help it is possible to efficiently increase the quality of the classification measured, for example, with the help of the accuracy. Due to the fact that the Choquet integral is resistant to data fluctuations and takes into account the quality (significance) of the information source, it appears to be an effective proposition for the final determination of what data, or more precisely, which records can be considered anomalous. The innovative approach to analyze transport data has not been used before. This article considers four publicly available databases covering different fields of application of transport systems. In a series of comprehensive numerical experiments, the Choquet integral-based approach has proven high efficiency for each of them. Moreover, we made a comparative analysis of the solutions before applying the Choquet integral and the results after its application.

Modern medicine has been increasingly using information technologies and computer systems to improve decision-making processes. Important examples of such activities are the collection, processing, and analysis of oculographic data. The correct interpretation of the measurement results plays a very important role here. Medical examinations based on eye-tracking allow for early detection of many diseases or disfunctions of human brain. Moreover, it can be seen that implementation of such methods improves human-computer interfaces. In this study, we propose an innovative solution based on ten approaches of anomaly detection with the use of fuzzy aggregation of their results. Also, our experiment is supported by the analysis of obtained anomaly scores by the specialists in the field of medicine. The results of the automatic and expert evaluation show high potential of our method. There are also some differences in the perception of the anomaly by machine learning techniques and experts’ judgement. Hence, in this paper we try to understand and explain them.

Modern information technologies allow for the collection, processing, and data analysis. A very important role of these systems can be observed in the analysis of medical records, particularly in the analysis of motion capture data. Detection and interpretation of data collected from movement recording devices enables for a fast diagnosis, e.g. disease or misfunction. Moreover, it can be assistive in the introduction of appropriate treatment or rehabilitation. Anomaly detection methods play a key role in the evaluation of this type of medical research results. In this study, we introduce an innovative approach based on the aggregation of the results of eleven anomaly detection classifiers outcomes with the fuzzy Choquet integral. Furthermore, the results of numerical experiments are confronted with the assessments of the experts in medical field. The results show the great potential of our method in supporting the decision-making process based on the motion capture data analysis. Moreover, we have caught the differences between the expert's understanding of anomaly and the anomalies in data found by the modern machine learning methods.

In this study, we present an innovative approach to deriving an aggregate classification score based on multiple classifiers based on generalizations of the Choquet integral. These generalizations are inspired by the quadratures known from numerical analysis, used to calculate integrals, e.g. the Newton-Cotes formula. The previous formulas for calculating generalizations of the Choquet integral used two (e.g. the case of the difference of these values), or three adjacent values related to the degrees of belonging of a given element to individual classes, related to one density of the fuzzy measure. In this article, we offer an interesting generalization. The novel enhancement is based on the replacement of typical product or t-norm appearing under the integral sign by forms related to mathematical quadratures. The formulas become more precise and better reflect the idea of integration. Moreover, a series of numerical experiments confirmed the advantage of the new approach over the existing ones.

Anomaly detection in data sets is one of the most important challenges for modern analysts and data administrators. It is usually based on algorithms that use raw data. In this study, we analyze the possibilities of improving the well-known Isolation Forest algorithm based on binary search trees for data preprocessing using the grouping of both attributes first, and then records within attribute groups. Attribute clustering is based on hierarchical grouping, while record grouping uses K-Means and Fuzzy C-Means. To describe the relationships between records, data membership functions are also used, built on the basis of record distances from centroids. This approach gives a new look at the possibilities of the Isolation Forest method and leads to a significant improvement in the results for selected public databases.