The number of cyberattacks has increased at an alarming rate in recent years, and cybersecurity is becoming a priority for many companies. This article aims to present a solution based on text classification for managing information related to cyber attacks and to facilitate the information and education of both regular users and IT specialists. In the experiments that will be described, a dataset containing articles related to the field of cybersecurity extracted from the Hacker News platform was used. The obtained results can serve as a starting point in the development of a Large Language Model (LLM) system to facilitate the search and dissemination of informational content.

The primary objective of this article is to enhance community safety by employing unsupervised machine learning algorithms in the development of a mobile application that provides accessible information on crime risks in specific areas for the users. By allowing individuals to avoid high-risk zones or take precautionary measures, the application aims to stimulate local economic activities such as tourism and real estate in safer areas. In the development of the application, the following unsupervised machine learning algorithms were used: K-Means, DBSCAN, and Hierarchical Clustering, along with an examination of environmental and social factors that contribute to urban crime. The experimental chapter presents the process of translating data into risk zones by comparing the performance of the three clustering algorithms. The algorithm with the best performance is integrated into the mobile application to ensure optimal functionality.

Light pollution has become a significant issue in recent years, drawing attention to its adverse impact not only on the environment, but also on our daily lives. It is crucial to implement measures to mitigate these negative effects by minimizing light pollution in areas where it has experienced exponential growth. Currently, there are no algorithms available that can determine light pollution from ground images. Therefore, the objective of this article is to classify night sky photos according to the Bortle scale. The proposed approach achieved promising results, with 99% accuracy in the Classification by light pollution. These results open new horizons for addressing light pollution and provide motivation for future research in this field.

The primary objective of this research paper is to conduct a real-life case study, centred around a substantial dataset procured from numerous Romanian public procurement tenders. The study aims to classify the bidders according to their compatibility and suitability, an assessment determined by the accuracy of different machine learning models. The dataset used, is a vast body of information. More precisely, it includes data from a total of 289,472 enterprises, 47,974 contracting authorities and 42,474 tenders. The use of this extensive data set ensures a robust and comprehensive analysis, thus enabling the extraction of meaningful insights and facilitating the drawing of reliable conclusions. In the quest to rank bidders effectively, the performance of nine intelligent algorithms is evaluated and compared. The top-performing algorithms in this context appear to be Decision Trees, along with ensemble methods derived from them, namely the Random Forest Classifier and the Extra Trees Classifier. The outcomes of this study are satisfying, outperforming the metrics of a similar study conducted in the same domain. Given the potential concerns about Romania’s integrity in dealing with contracting companies for public procurement auctions, this study holds significant value in its ability to validate the decision-making process employed in previous auctions. By casting light on the effectiveness of past procurement decisions and offering strategic guidance for future acquisitions, this research holds substantial implications for improving transparency and streamlining the bidder selection process within the public procurement area. Therefore, with potential for future work upon it, it promises to become a valuable addition in enhancing the integrity and efficiency of public procurement in Romania.

Weather effects are long known to significantly impact human health, especially in the context of climate change and their associated extreme weather conditions. Both healthcare systems and individuals are undoubtedly negatively affected by these phenomena. Thus, our study takes a dual approach towards leveraging Machine Learning for adaptation and prevention. The proposed solution consists of a regression model for hospital admissions and a classification model for risk prediction considering weather conditions and comorbidities for three diseases: heart failure, cerebral infarction and respiratory failure. We relied on meteorological data from official weather stations in Romania and hospitalization data from the five most populated Romanian cities (Bucharest, Iasi, Constanta, Cluj-Napoca, and Timisoara). We analyzed the effectiveness of various Machine Learning algorithms for both tasks. To evaluate the influence of environmental and health factors on critical incidents, we explored three experimental configurations: all features, excluding weather parameters, and excluding medical information. For both demand forecasting and risk evaluation, Extreme Gradient Boosting delivered the best metrics upon the analyzed features: 11.2% mean absolute percentage error for hospital admissions forecasting and recall of 94.1% for heart failure, 86.5% for cerebral infarction, and 65.1% for respiratory failure. The main conclusion is that selected weather parameters considered have no major significance for the prediction outputs for the selected diseases, especially in comparison with other features such as geographical and temporal information in case of admission prediction, and age, pre-existing medical conditions of an individual in case of disease risk estimation. 1

This paper presents an approach based on optimized data search and data analysis of high dimensional clinical health records, for the purpose of identifying the patients with significant risk of fracture to be introduced to the Fracture Liaison Service (FLS) for further investigations and treatment. The main objective of the work is to speed up and automate the process of identifying the patients suitable for FLS and provide a real decision support for the clinical team by fast information retrieval and structuring. The performed analysis and the current results are promising and support our research aim for the optimization of the process in the real life scenario.