Detecting early signs of anomalous or fraudulent activity in blockchain transaction networks remains a challenging task. Most existing approaches rely on static graph representations or supervised methods, overlooking the temporal evolution of structural changes. Through this research, we propose an unsupervised framework for identifying emerging structural changes in Bitcoin transaction graphs using topological drift. The blockchain is modeled as a sequence of temporal graphs, where node-level topological features are tracked across consecutive time windows. We introduce the Topological Volatility Index (TVI) to quantify global structural instability over time. Experiments on real-world Bitcoin data show that the network is generally stable, with occasional high-volatility spikes corresponding to significant structural shifts. These events are driven by a small number of nodes with high drift, enabling the identification of distinct structural patterns. Our results demonstrate that topological drift can serve as an effective early indicator of behavioral changes, supporting unsupervised anomaly detection and forensic analysis.

This paper presents a multi-agent system MAS-PatientCare, specifically designed to manage patient scheduling, resource allocation and diagnostic processes in hospitals. The system architecture integ-rates specialized agents, each with different responsibilities, including patient management, hospital resource allocation, scheduling, medical specialization, data preprocessing, machine learning, and decision support, to improve operational efficiency and quality of patient care across multiple hospital departments, including colonoscopy and emergency, with the flexibility to add more departments as needed. We explore the interconnectivity and collaboration between these agents, detailing how they interact to ensure seamless operations. To evaluate the effectiveness and practical applicability of the MAS-PatientCare system, it has been tested in two different situations: in emergency situations in hospitals and in the screening process for patients undergoing a colonoscopy. The impact of the MAS-PatientCare system on patient outcomes, particularly in terms of reducing waiting times, improving diagnostic accuracy and improving resource utilisation is analysed.

With the advent of genomics and next generation sequencing, metagenomics has become a discipline in its own right, enabling for the first time the study of complex microbial ecosystems containing species that cannot be cultured in the lab. Sequencing and characterizing these microbial genomes enables the identification of novel antimicrobial resistance genes (ARGs) and biosynthetic gene clusters (BGCs), with potential applications in industry. De novo sequence assembly, as a crucial step in metagenomics, is currently being implemented with tools such as MEGAHIT, which employ a multiple $k$ -mer approach, where several intermediate assemblies are built from k-mers from a range of sizes and merged at the end. The objective of this study was to investigate the impact that this range of k-mer sizes could have on a range of quality metrics, including the number of identified ARGs and BGCs, using a series of assembly scenarios involving 7 samples collected from the Movile Cave in Romania We found that larger k-mer sizes and more fine-grained ranges of intermediate steps were associated with increased metrics, including larger numbers of identified BGCs. Future studies are required in order to better understand the role of the range of k-mer sizes in MEGAHIT in the context of microbial communitywide identification of ARGs and BGCs.

This paper proposes an innovative approach to achieving real-time polymorphic behavior detection, and its direct application to blockchain-focused smart-contracts. We devise a method based on a non-deterministic finite state machine to perform approximate pattern-matching, using a look-ahead mechanism implemented through a concept similar to that of a sliding window, and using threshold-based similarity checking at every state in the automaton. We introduce and formalize our approach, discuss the challenges we faced and then test it in a real-world environment. The experimental results obtained showed a significant speed-up of our approach, as compared to the classic similarity measures used commonly in such scenarios.

The use of short-term prediction techniques with the purpose of anticipating hyper- or hypoglycemic episodes for patients with type 1 diabetes has gained a lot of interest in the past years. Both hypoglycemia and hyperglycemia have a major impact and considerable effects in the life of a person with diabetes. The prediction of abnormal values of the blood glucose is a vital step in improving the life quality of a diabetic patient. This paper focuses on developing models for predicting blood glucose levels, to warn type 1 diabetes patients about potential hyper- or hypoglycemic events. Our main objective is to help the daily management of people with type 1 diabetes and to improve their long-term health condition. We will compare four models: a regression model, recurrent neural networks, Long Short-Term Memory, and Gated Recurrent Unit.

This study explores the application of Long ShortTerm Memory algorithms to predict blood glucose levels, leveraging sequence-based data from a Dexcom G6 Continuous Glucose Monitoring sensor. This research methodically details a rigorous data cleansing process that transforms raw glucose readings into meaningful, sequenced input for Long Short-Term Memory models. By feeding the Long Short-Term Memory with structured data in a variety of volumes, it was analyzed and discussed how the quantity of information influences the accuracy of glucose level predictions. This investigation holds significant implications for the advancement of personalized glycemic control, potentially improving the dayto-day management and overall quality of life for individuals living with diabetes.

Gastrointestinal polyps are common abnormalities detected during colonoscopy screenings, and accurate classification of these polyps is essential for effective diagnosis and treatment planning. This study aims to explore machine learning techniques’ performance for classifying gastrointestinal polyps in colonoscopy video clips. A comparative analysis is conducted to evaluate the performance of seven machine learning algorithms, including three types of tree-based algorithms, three types of boosting algorithms, and two function-based algorithms. The dataset consists of 152 instances, encompassing hyperplastic, serrated, and adenoma lesions, with a total of 76 polyps. The results of this study provide insights into the effectiveness of machine learning algorithms for improving the efficiency and accuracy of gastrointestinal polyp classification, ultimately contributing to enhanced patient care and diagnostic outcomes.

We are presenting a fast and innovative approach to performing approximate pattern-matching for plagiarism detection, using an NDFA-based approach that significantly enhances performance compared to other existing similarity measures. We outline the advantages of our approach in the context of blockchain-based non-fungible tokens (NFTs). We present, formalize, discuss and test our proposed approach in several real-world scenarios and with different similarity measures commonly used in plagiarism detection, and observe significant throughput enhancements throughout the entire spectrum of tests, with little to no compromises on the accuracy of the detection process overall. We conclude that our approach is suitable and adequate to perform approximate pattern-matching for plagiarism detection, and outline research directions for future improvements.

We are presenting a fast and innovative approach to performing approximate pattern-matching for plagiarism detection, using an NDFA-based approach that significantly enhances performance compared to other existing similarity measures. We outline the advantages of our approach in the context of blockchain-based non-fungible tokens (NFTs). After testing in real-world scenarios, we conclude that our approach is suitable and adequate to perform approximate pattern-matching for plagiarism detection, yet significantly faster and therefore more suitable for big data analysis.

We are introducing a new computational model for heterogeneous architectures for performing enhanced security validations and processing of transactions on-chain, a model applicable to both permissioned and permissionless blockchains. By proposing an efficient storage model for the Merkle trees involved in the transaction and security validation processing, and combining it with an efficient heterogeneous approach for processing it, we are outlining the improved computational advantages to ensuring attack resilience on-chain, which surpasses the performance of traditional CPU-based approaches by a significant factor.

Vital sign monitoring outside the clinical environment based on wearable sensors ensures better support in assessing a patient's health condition, and in case of health deterioration, automatic alerts can be sent to the care providers. In everyday life, the users can perform different physical activities, and considering that vital sign measurements depend on the intensity of the activity, we proposed an architecture based on the multi-agent paradigm to handle this issue dynamically. Different types of agents were proposed that processed different sensor signals and recognized simple activities of daily living. The system was validated using a real-life dataset where subjects wore accelerometer sensors on the chest, wrist, and ankle. The system relied on ontology-based models to address the data heterogeneity and combined different wearable sensor sources in order to achieve better performance. The results showed an accuracy of 95.25% on intersubject activity classification. Moreover, the proposed method, which automatically extracted vital sign threshold ranges for each physical activity recognized by the system, showed promising results for remote health status evaluation.

Decoding the human genome in the past decades has brought into focus a computationally intensive operation through DNA profiling. The typical search space for these kinds of problems is extremely large and requires specialized hardware and algorithms to perform the necessary sequence analysis. In this paper, we propose an innovative and scalable approach to exact multi-pattern matching of nucleotide sequences by harnessing the massively parallel computing power found in commodity graphical processing units. Our approach places careful consideration on preprocessing of DNA datasets and runtime performance, while exploiting the full capabilities of the heterogeneous platform it runs on. Finally, we evaluate our models against real-world DNA sequences.

The complexity of business environment often forces sales representative to take decisions using intuition and/or subjective models created through experience instead of data driven decisions. However, data driven decisions generate better revenue. In this work, we present a methodology to create a Decision Support System (DSS) for Business to Business (B2B) sales that generates recommendations for actions to follow and actions to avoid to increase the chances of successfully closing a deal. At the core of our proposed DSS, is a three-step methodology relying on a particular approach of using belief rules. In the first step we generate, based on available data on previous won/lost deals, some forecasting models. As the best performing models are of black box type, the second step aims to extract if-then conventional rules that are converted, in the last step, into belief rules. Rather than using an attribute value probability matching to activate a belief rule, we propose a cost of activation that takes into account the current state of the active deal and the cost of transitioning the active deal's attributes into a new state that could activate a belief rule. Furthermore, we propose and analyze two methods of calculating the cost of transition related to the effort the sales representative should dispose to move the current deal into the new state. The three-step approach was used to create a DSS for a real world B2B complex sales Customer Relationship Management Systems and the results are discussed and included in this paper.

In today's world of big data, multi-tenant cloud Customer Relationship Management Systems with an ever increasing pressure to customize offerings for each prospect, there is a need for distributed problem solving that can help salespeople win sales using data driven recommendations. With this work we are proposing a Decision Support System constructed as a multi-agent architecture for Business to Business Customer Relationship Management Systems that can guide salespeople during the conversion of a prospect into a client. The implementation and the results using real-world CRM data are presented and discussed in this paper.

In this work, we are proposing a methodology for data-driven decision making using sentiment analysis. The analysis of sentiment is done by text mining the activity notes recorded in Customer Relationship Management Systems used to manage complex sales in business to business environments. We built the sentiment enhanced sales prediction models using Artificial Neural Networks, Support Vector Machines and Random Forests and involving different sentiment features. The approach produced meaningful results with Random Forest obtaining the best improvement compared to a baseline model without sentiment features. The best model showed that new attributes incorporating sentiment information improved the accuracy from a baseline of 85.15% to 89.11 %. This model was used to conduct an analysis and an evaluation of the steps needed to be taken to win a possible losing deal in a real-world business to business customer relationship management system.

We are presenting a new architectural model for enhancing blockchain processing of transactions and security validation through heterogeneous computing, by focusing and improving the consensus validation process, specifically the Merkle tree implementation and validation, both in terms of processing and storage efficiency. By using a heterogeneous architecture, coupled with an efficient storage model for instant transfers of data between memory and video memory, the validation process efficiency can be improved and offloaded from the CPU, with a direct result in increased transactional speeds and enhanced security.

We are presenting a highly-efficient, novel architecture (which we call FAST, or Forensic Analysis of Sensitive Traces) for high-performance big data forensics for heterogeneous systems (CPU and GPU-based). Our model uses a highly-compact storage format of the widely known Aho-Corasick algorithm [1], as well as a partial pruning mechanism to ensure the lowest possible memory footprint, while maximizing throughput performance. We are comparing our performance with classic methods used in data forensics and observe significant memory footprint improvements, as well as massive throughput improvements throughout all stages of big data processing.

With the ever-increasing capabilities of modern hardware and breakthroughs in the DNA biometrics field, we are presenting a new, scalable and innovative method to accelerate the DNA analysis process used in criminal investigations, by building an improved methodology for using large-scale GPU-based automata for performing high-throughput pattern-matching. Our approach focuses on all important stages of preparing for the pattern-matching process, tackling with all major steps, from creation, to preprocessing, to the runtime performance. Finally, we experiment using real-world DNA sequences and apply the process to the human DNA genome, for an evaluation of our implementation.

In this paper we present a concept project for a self developing system based on agents built for a hospital. The system monitors patients during and after being released from hospitalization, with the aim of understanding patterns and predicting future problems. Due to its complexity and dynamism the agents must be automatically generated. They need to cooperate and "compete" with each other in order to get good results. By combining meta-heuristic algorithms with reinforcement and clustering techniques we target a large degree of autonomy in decision making.

We are proposing a new, memory-efficient approach to optimizing DNA pattern-matching in bioinformatics through a heterogeneous implementation and new architectural layout, that poses several advantages over usual approaches, which we discuss in detail. We are applying our approach on a subset of DNA sequences part of the FASTA open database, under different hardware settings, and observe a significant performance increase in our heterogeneous implementation. With a practical reduction of 23 times less memory usage than a classic implementation of the same algorithm, and massive scaling capabilities for high-throughput DNA-matching, our approach proves its feasibility for scalable heterogeneous architectures.