@upt.ro
Communications Department / Faculty of Electronics, Telecommunications, and Informational Technologies
Politehnica University Timisoara
Computer Science, Artificial Intelligence, Information Systems, Electrical and Electronic Engineering
Scopus Publications
Scholar Citations
Scholar h-index
Scholar i10-index
Marian Bucos and Bogdan Drăgulescu
MDPI AG
Misinformation poses a significant challenge in the digital age, requiring robust methods to detect fake news. This study investigates the effectiveness of using Back Translation (BT) augmentation, specifically transformer-based models, to improve fake news detection in Romanian. Using a data set extracted from Factual.ro, the research finds that BT-augmented models show better accuracy, precision, recall, F1 score, and AUC compared to those using the original data set. Additionally, using mBART for BT augmentation with French as a target language improved the model’s performance compared to Google Translate. The Extra Trees Classifier and the Random Forest Classifier performed the best among the models tested. The findings suggest that the use of BT augmentation with transformer-based models, such as mBART, has the potential to enhance fake news detection. More research is needed to evaluate the effects in other languages.
Marian Bucos and Georgiana Țucudean
MDPI AG
This paper aims to investigate the use of a Romanian data source, different classifiers, and text data augmentation techniques to implement a fake news detection system. The paper focusses on text data augmentation techniques to improve the efficiency of fake news detection tasks. This study provides two approaches for fake news detection based on content and context features found in the Factual.ro data set. For this purpose, we implemented two data augmentation techniques, Back Translation (BT) and Easy Data Augmentation (EDA), to improve the performance of the models. The results indicate that the implementation of the BT and EDA techniques successfully improved the performance of the classifiers used in our study. The results of our content-based approach show that an Extra Trees Classifier model is the most effective, whether data augmentation is used or not, as it produced the highest accuracy, precision, F1 score, and Kappa. The Random Forest Classifier with BT yielded the best results of the context-based experiment overall, with the highest accuracy, recall, F1 score, and Kappa. Furthermore, we found that BT and EDA led to an increase in the AUC scores of all models in both content-based and context-based data sets.
Mihai Ursan and Marian Bucos
IEEE
This study examines the student admission process at Politehnica University Timisoara's Faculty of Electronics, Telecommunications, and Information Technologies for master's engineering programs. The focus of the study is the student's admission prediction into the mainstream programs offered by our faculty. The programs in this category are the most successful and stable. Most of the features we use have been geared toward the options of future degree programs students may choose to pursue. In these experiments, we used several evaluation methods, including stratified cross-validation and evaluation of unseen data chunks that were not used for training. The Gradient Boost Classifier performs the best of the six models. Among the conclusions drawn is that secondary options do not provide relevant information for predictive models. We implemented the experiments using a low-code framework; as a result, the predictive analysis pipeline was largely automated.
Georgiana Tucudean and Marian Bucos
IEEE
Anomalies and fake data can be identified by applying supervised learning algorithms to news sources. These techniques can help reduce the negative impact of fake news on consumers. Fake news is a widespread problem around the world and is also gaining momentum in Romania. For the detection of fake news in Romanian, we investigate methods to construct a Romanian data set and apply algorithms that offer the highest performance. To improve performance, we use a data augmentation technique called back translation in conjunction with the Support Vector Machine classifier.
Georgiana Simion, Catalin Daniel Caleanu, Marian Bucos, and Bogdan Dragulescu
IEEE
The creation of agricultural clusters is one of the methods by which farmers can increase their competitiveness and generate economic growth. Machine learning can help to decide which are the best clusters to form in an area. The aim of the paper is to examine the possibility of using clustering on an agricultural platform. This platform is under development, and currently there are no real data available for training purposes. Therefore, we constructed a data set by integrating multiple data sources. This was constructed to be as similar as possible to the characteristics of the data available in the final platform. The proposed data set consists of 4480 records and 60 features, ranging from geographical location to crop yield. Next, the experimental part proposes a scenario for grouping the farmers based on geographic proximity and crop yield.
Marian Bucos and Bogdan Dragulescu
IEEE
The present work considers the possibility of using Moodle course logs and student performance indicators within the Database Systems course to apply the K-Means clustering algorithm. Clusters of students are identified and explained to partition students with similar study behaviours and performance. Moreover, the understanding of the five groups emerged in cluster analysis allowed us to identify a cluster that contains 86% of students at risk of not completing the course activity. One important aspect that differentiates our study from other similar works is the use of data collected over a long period of time, from 2015 to 2019. The final data set, obtained after preprocessing, contains no less than 185.206 course logs.
Bogdan Drăgulescu and Marian Bucos
Springer International Publishing
Bogdan Drăgulescu, Marian Bucos, and Radu Vasiu
Springer International Publishing
Bogdan Dragulescu, Marian Bucos, and Radu Vasiu
Faculty of Electrical Engineering and Computing, Univ. of Zagreb
The increased usage of information technologies in educational tasks resulted in high volume of data, exploited to build analytical systems that can provide practical insight in the learning process. In this paper, we propose a method of running social network analysis on multiple data sources (academic years, communication tools). To achieve this, the collected data that describe social interactions were converted into a common format by employing a prior developed semantic web educational ontology. Using a mapping language the relational data set was linked to the appropriate concepts defined in the ontology and then it was exported in RDF format. The means for SPARQL access was also provided. Subsequently, query patterns were defined for different social interactions in the educational platform. To prove the feasibility of this approach, Gephi tool set was used to run SNA (Social Network Analysis) on data obtained with the SPARQL queries. The added value of this research lies in the potential of this method to simplify running social network analysis on multiple data sets, on a specific course or the entire academic year, by simply modifying the query pattern.
Radu Vasiu, Diana Andone, Andrei Ternauciuc, and Marian Bucos
IEEE
In recent years the higher education system in Romania has been through a series of transformations. These transformations reflects European tendencies but also the changes due to the new Romanian Education Law. The universities involved in this project, described here, are at the avant-garde of these changes and promote innovation in post-graduate education. But are the prospective students and the universities ready to take the challenge? And is it the eLearning environment going to be a viable model for a multi-regional Masters degree? A study, which focuses on the first semester of piloting, was conducted recently in order to answer these questions.
B. Dragulescu, I. Ermalai, M. Bucos, and M. Mocofan
IEEE
Extraction of information relays heavily on the user. The correct extraction of information is very important especially in an eLearning environment. We can ensure correct distribution of important data (contact information, calendar events, and locations) by encapsulating that information in the correct formats and rely on automated agents for data processing. In this paper, we propose a model of republishing the user contact information from Moodle in hCard and vCard formats, and also QR codes for mobile tagging. This application model can be extended in the future to data regarding calendar events (vcalendar, hcalendard, ical) and locations.
M. Mocofan, I. Ermalai, M. Bucos, M. Onita, and B. Dragulescu
IEEE
For indexing an image databases there are used many features. In this paper are used features like histograms for the color components, stochastic moments, few parameters of the co-occurrences matrix and the Gabor filters decomposition. Using all of them we obtain good performance in the field of content-based image indexing and retrieval. There are some speed problems because are many features used in the search process. Here, I present an algorithm for indexing and retrieval, which use a supervised tree algorithm for speed increasing. The area of applications is very wide: multimedia documents, transaction systems, medical application.
Marian Bucos, Bogdan Dragulescu, and Marius Veltan
IEEE
The educational ontology we presented in this paper is used to model web-based e-learning systems for higher education. Our main objective was to create domain ontology to play an important role in representing higher education concepts and to assist specialized e-learning systems. We have stressed the need for developing educational models that meet the expectations of higher education community with regards to e-learning adaptation and efficiency by employing ontologies and Semantic Web techniques.
M. Mocofan, R. Vasiu, M. Bucos, M. Ionita, and I. Ermalai
IEEE
The modernization of the buildings with the most performing technologies starts to grow more and more, and like a consequence, many people appeal to this technique, to live in a more safety house, and why not, more comfortable. We propose them a multimedia interface connected with a relational database to control the scenarios of smart buildings. The scenarios include all the settings for the house (lights, temperature and music), for example "romantic dinner".
R. Vasiu, N. Robu, D. Andone, and M. Bucos
IEEE
In recent years the higher education system in Romania has been through a series of transformations. These transformations reflects European tendencies but also the growth of the Romanian economy. The changes affect also distance education, which is the actual form of delivering several specializations, as off-campus solutions. The "Politehnica" University of Timisoara is at the avant-garde of these changes and promotes innovation in education. E-learning is seen as a further step in high-education delivery. But are the prospective students and the universities ready to take the challenge? And is it e-learning going to be a viable model for an International Masters degree in Multimedia? A study, which focuses on the historic and economic development, legislation, and ICT penetration, was conducted recently in order to answer these questions.