Michael Paraskevas

Prof. Dr. Michael Paraskevas

Electrical and Computer Engineering Department, University of Peloponnese, 221 00 Tripoli, Greece

Research Focus & Metrics

Academic Career Length

7Years (2020-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
21 October 2020 in Applied Sciences
Verified
Journal Article
Open access
PDF file
IMIosif Mporas
Iosif Mporas
Isidoros PerikosIsidoros Perikos
Isidoros Perikos
VKVasilios Kelefouras
Vasilios Kelefouras
Michael ParaskevasMichael Paraskevas
Michael Paraskevas
In this article, we present a framework for automatic detection of logging activity in forests using audio recordings. The framework was evaluated in terms of logging detection classification performance and various widely used classification methods and algorithms were tested. Experimental setups, using different ratios of sound-to-noise values, were followed and the best classification accuracy was reported by the support vector machine algorithm. In addition, a postprocessing scheme on decision level was applied that provided an improvement in the performance of more than 1%, mainly in cases of low ratios of sound-to-noise. Finally, we evaluated a late-stage fusion method, combining the postprocessed recognition results of the three top-performing classifiers, and the experimental results showed a further improvement of approximately 2%, in terms of absolute improvement, with logging sound recognition accuracy reaching 94.42% when the ratio of sound-to-noise was equal to 20 dB.

6 January 2020 in Entropy
Verified
Journal Article
Open access
PDF file
PSPascal Schirmer
Pascal Schirmer
IMIosif Mporas
Iosif Mporas
Michael ParaskevasMichael Paraskevas
Michael Paraskevas
In this article an energy disaggregation architecture using elastic matching algorithms is presented. The architecture uses a database of reference energy consumption signatures and compares them with incoming energy consumption frames using template matching. In contrast to machine learning-based approaches which require significant amount of data to train a model, elastic matching-based approaches do not have a model training process but perform recognition using template matching. Five different elastic matching algorithms were evaluated across different datasets and the experimental results showed that the minimum variance matching algorithm outperforms all other evaluated matching algorithms. The best performing minimum variance matching algorithm improved the energy disaggregation accuracy by 2.7% when compared to the baseline dynamic time warping algorithm.

9 March 2023 in Information
Verified
Journal Article
Open access
PDF file
SLSpyridon Louvros
Spyridon Louvros
Michael ParaskevasMichael Paraskevas
Michael Paraskevas
Theofilos ChrysikosTheofilos Chrysikos
Theofilos Chrysikos
Fifth-generation and more importantly the forthcoming sixth-generation networks have been given special care for latency and are designed to support low latency applications including a high flexibility New Radio (NR) interface that can be configured to utilize different subcarrier spacings (SCS), slot durations, special scheduling optional features (mini-slot scheduling), cloud- and virtual-based transport network infrastructures including slicing, and finally intelligent radio and transport packet retransmissions mechanisms. QoS analysis with emphasis on the determination of the transmitted packets’ average waiting time is therefore crucial for both network performance and user applications. Most preferred implementations to optimize transmission network rely on the cloud architectures with star network topology. In this paper, as part of our original and innovative contribution, a two-stage queue model is proposed and analytically investigated. Firstly, a two-dimension queue is proposed in order to estimate the expected delay on priority scheduling decisions over the IP/Ethernet MAC layer of IP packet transmissions between gNB and the core network. Furthermore, a one-dimension queue is proposed to estimate the average packet waiting time on the RLC radio buffer before being scheduled mainly due to excessive traffic load and designed transmission bandwidth availability.

29 April 2025 in Computers
Verified
Journal Article
Open access
PDF file
KKKonstantinos Kyritsis
Konstantinos Kyritsis
CLCharalampos M Liapis
Charalampos M Liapis
Isidoros PerikosIsidoros Perikos
Isidoros Perikos
Michael ParaskevasMichael Paraskevas
Michael Paraskevas
VKVaggelis Kapoulas
Vaggelis Kapoulas
This study conducts an in-depth investigation of the performance of six transformer models using 12 different datasets—10 with three classes and two with two classes—on sentiment classification. We use these six models and generate all combinations of triple schema ensembles, Majority and Soft vote. In total, we compare 46 classifiers on each dataset and see in one case up to a 7.6% increase in accuracy on a dataset with three classes from an ensemble scheme and, in a second case, up to 8.5% increase in accuracy on a dataset with two classes. Our study contributes to the field of natural language processing by exploring the reasons for the predominance, in this particular task, of Majority vote over Soft vote. The conclusions are drawn after a thorough investigation of the classifiers that are co-compared with each other through reliability charts, analyses of the confidence the models have in their predictions and their metrics, concluding with statistical analyses using the Friedman test and the Nemenyi post-hoc test with useful conclusions.

Verified
Journal Article
Open access
PDF file
CLCharalampos M Liapis
Charalampos M Liapis
KKKonstantinos Kyritsis
Konstantinos Kyritsis
Isidoros PerikosIsidoros Perikos
Isidoros Perikos
NSNikolaos Spatiotis
Nikolaos Spatiotis
Michael ParaskevasMichael Paraskevas
Michael Paraskevas
The present study explores the field of text classification in the Greek language. A novel ensemble classification scheme based on generated embeddings from Greek text made by the multilingual capabilities of the E5 model is presented. Our approach incorporates partial transfer learning by using pre-trained models to extract embeddings, enabling the evaluation of classical classifiers on Greek data. Additionally, we enhance the predictive capability while maintaining the costs low by employing a soft voting combination scheme that exploits the strengths of XGBoost, K-nearest neighbors, and logistic regression. This method significantly improves all classification metrics, demonstrating the superiority of ensemble techniques in handling the complexity of Greek textual data. Our study contributes to the field of natural language processing by proposing an effective ensemble framework for the categorization of Greek texts, leveraging the advantages of both traditional and modern machine learning techniques. This framework has the potential to be applied to other less-resourced languages, thereby broadening the impact of our research beyond Greek language processing.

6 September 2024 in Smart Cities
Verified
Journal Article
Open access
PDF file
IDIoannis Daramouskas
Ioannis Daramouskas
Isidoros PerikosIsidoros Perikos
Isidoros Perikos
Michael ParaskevasMichael Paraskevas
Michael Paraskevas
Vaios LappasVaios Lappas
Vaios Lappas
VKVaggelis Kapoulas
Vaggelis Kapoulas
LoRa technology is a recent technology belonging to the Low Power and Wide Area Networks (LPWANs), which offers distinct advantages for wireless communications and possesses unique features. Among others, it can be used for localization procedures offering minimal energy consumption and quite long-range transmissions. However, the exact capabilities of LoRa localization performance are yet to be employed thoroughly. This article examines the efficiency of the LoRa technology in localization tasks using Time Difference of Arrival (TDoA) measurements. An extensive and concrete experimental study was conducted in a real-world setup on the University of Patras campus, employing both real-world data and simulations to assess the precision of geodetic coordinate determination. Through our experiments, we implemented advanced localization algorithms, including Social Learning Particle Swarm Optimization (PSO), Least Squares, and Chan techniques. The results are quite interesting and highlight the conditions and parameters that result in accurate LoRa-based localization in real-world scenarios in smart cities. In our context, we were able to achieve state-of-the-art localization results reporting localization errors as low as 300 m in a quite complex 8 km × 6 km real-world environment.