Keynote Speakers
![](images/durresi.jpg)
Prof. Arjan Durresi
Biography
Arjan Durresi is a Professor of Computer Science at Indiana University Purdue University in Indianapolis, Indiana. In the past, he held positions at LSU and The Ohio State University. His research interests include trustworthy decision making and AI, networking, and security. He has published about 100 articles in journals and over 200 articles in conference proceedings and seven book chapters. He also has authored over thirty contributions to standardization organizations such as IETF, ATM Forum, ITU, ANSI and TIA.
![](images/sacile.jpg)
Prof. Roberto Sacile
Biography
Roberto Sacile (Member, IEEE) received the Laurea degree in electronic engineering from the University of Genova, Italy, in 1990, and the Ph.D. degree, in 1994. Since 2000, he has been with the University of Genova, where he is currently a Professor of automation and systems engineering. His research interests include systems engineering and its application to the transportation, logistics, and energy sectors. More recently, he has been working in human machine interfaces to enhance safety in the automotive sector.
![](images/simona.jpg)
Prof. Simona Sacone
Biography
Simona Sacone (Member, IEEE) received the master’s degree in electronic engineering and the Ph.D. degree in electronic engineering and computer science from the University of Genoa, Italy, in 1992 and 1997, respectively. She is currently an Associate Professor of automatic control with the Department of Informatics, Bioengineering, Robotics and Systems Engineering, University of Genoa, where she acts as a Coordinator of the Ph.D. course on systems engineering. She has authored or coauthored more than 150 papers published in international journals, international books, and international conference proceedings. She is the Chair of the Technical Committee on Planning and Control of Transportation Networks of the IEEE Intelligent Transportation Systems Society. She serves as an Associate Editor for the IEEE Transactions on Intelligent Transportation Systems and the IEEE Control Systems Magazine.
Proceedings ICITEE21
CONFERENCE PROCEEDINGS include only papers accepted for publication by ICITEE21 program committee, which meet all the formatting requirements and are accepted by the management of the corresponding section.
Proceedings book of ICITEE 21 |
ISBN 978-9928-329-52-3 is published by: Enti Botues “Gjergj Fishta” and submitted to:
- National Library of Albania;
- Library of the Academy of Sciences of Albania;
- Scientific Library of Polytechnic University of Tirana and other universities;
- Libraries of the programme committee university afiliations;
Session 1: Software Engineering and Application Development
Ina Papadhopulli, Realb Kushe - Challenges and Strategy Solutions for Continuous Software Engineering: A Systematic Literature Review
Faculty of Information Technology, Polytechnic University of Tirana, Tirana, Albania {ipapadhopulli, realb.kushe}@fti.edu.al
Abstract. The implementation of Continuous Software Engineering (CSE) practices is a must for software development companies to succeed in a highly competitive market. Software products are continuously released in small and incremental cycles, which turns to have its associated risks of failure. This systematic literature review investigates and classifies the challenges present in each stage of the Continuous Integration/Continuous Deployment (CI/CD) pipeline and the available strategies and tools used to address and solve these challenges. We analyzed forty-six relevant papers, filtered from four digital libraries, and provided information published in the last five years on behalf of CSE practices. Our study emphasizes that the selection of the right CSE strategy and automation tool to use is based on the project.Besjana Muraku, Evis Trandafili, Elinda Meçe - Agile Methods for Distributed Software Development: Literature Review
Faculty of Information Technology, Polytechnic University of Tirana, Tirana, Albania {besjana.muraku,etrandafili, ekajo}@fti.edu.al
Abstract. It is a trend nowadays for software projects to be developed by teams distributed geographically in different locations. Despite the benefits provided by this globalization, projectmanagement methods or softwaredevelopment processes must be adapted. Inthis paper we analyze the publications conducted in recent years to identify the challenges, trends,and the less explored area of Agile Methodologies for distributed software development (DSD). We have conducted aSystematic Literature Review (SLR) for the publications through the years 2017-2021 where we could give an answer to the questions: What evidence is provided for DSD? Which are the factors that impact agile development in distributed software development?and are there new models proposed or novel approaches for Agile in DSD?Results: We identified an increased interest in Agile for Distributed Software Development where in focus for the literature where the case studies and their feedback with guidelinesand proposal of hybrid models implementation. Scrum and Extreme Programming are being the most reviewed methods with an interest in Peer Programming and Lean Programming. At the same time, a lot of attention is given to scaled agile and its adoption for distributed teams too, where most of the analysis is focused on Scrum of Scrums, Scaled Agile Framework and Disciplined Agile Delivery.Evis Trandafili, Elinda Meçe, Artemisa Sinorukaj, Besjan Tomja, Ersi Bajramaj - A survey on frameworks used for IoT application development
Faculty of Information Technology, Polytechnic University of Tirana, Tirana, Albania {etrandafili,ekajo, artemisa.sinorukaj, besjan.tompja, ersi.bajramaj}@fti.edu.al
Abstract. Our current lifestyle has strongly embraced technologies that makeus able to control and hand lereal-world events. IoT system sare internet-enabled technologies and infrastructures that facilitate the collection of information from different intelligent devices that need to collaborate to provide fast and secure communication. Existing frameworks used for IoT application development, address several obstacles of IoT systems implementations and guarantee a reliable product with high overall quality. In this work we present a systematicliterature review about the usage of theframeworks thatease the IoT application development process in the aspect of programming and architectural design, security, and testing. We followed a predefined review protocol and identified a collection of frameworks. For each framework we present the main features, the impact in the facilitation of the development process for the developer, solutions it offers for addressing issues in IoT and its applications in real-life scenarios.Elma Zanaj, Ledjon Agalliu, Nikolin Ngjela, Elton Ngjela - Lekotech, an information ecosystem for automotive
Faculty of Information Technology, Polytechnic University of Tirana, Tirana, Albania {ezanaj, ledjon.agailliu}@fti.edu.al2 Lekotech, Tirana Albania {nikolinngjela, eltongjela}@thedancingbits.com
Abstract. Every part of our lives is being affected by information technology. Of course, different areas of life develop at different speeds when interacting with the digital world. One of the areas where digitalization is moving at a very slow pace, compared to other areas, is that of spare parts for automotive, especially those used. It is a whole profitable industry which loses significant amounts of income by not having accurate data on the products and services it deals with, supply chains, current innovations and updates, the flow of work processes, etc. This situation brings many difficulties for all actors involved in this industry, especially customers. Lekotech, is an information ecosystem, which in the essence of its existence, has the digitalization of this industry by completely revolutionizing the work processes, sales -customer interactions, the way of identifying the products that are sold, their tracking, determining their compatibility, etc.Session 2: Intelligent Systems and Machine Learning
Deivis Disha, Linda Senigagliesi, Ennio Gambi - Comparison of deep learning techniques forhuman activity classification based on featureextraction from range-Doppler maps
Universita Politecnica delle Marche, Ancona, Italy d.disha@pm.univpm.it, l.senigagliesi@staff.univpm.it, e.gambi@univpm.it
Abstract. Human activity classification is assuming a great relevance in many fields,including the well-being of the elderly. Many methodologies to improve the predictionof human activities, such as falls or unexpected behaviors, have been proposed over theyears, exploiting different technologies. In this paper, we consider a dataset containingdifferent types of activities related to human walking. We started by processing rawdata obtained from an automotive radar and then we perform feature extraction fromrange-Doppler maps. To obtain reliable results we compare different deep learningarchitectures and different optimizers. We show through results that an accuracy ofmore then 97% is achieved with an appropriate selection of the network parameters.Enida Sheme, Alba Merdani - Smart Environment IoT Project for Air Quality Monitoring and Prediction using Machine Learning
Faculty of Information Technology, Polytechnic University of Tirana, Tirana, Albania {esheme,alba.merdani}@fti.edu.al
Abstract. A friendly environment and good quality of air are key factors for a healthy life. Air pollution has become a serious problem in Albania as well. Technology is the right tool, that can help to monitor and analyze environmental air quality parameters. One of today's challenges with technology is process automation and improvement by monitoring, controlling,and analyzing data through artificial intelligence. To facilitate, improve and secure work processes and community living, more and more studies and projects are being implemented in the field of IoT, intelligent and automated systems. In this paper, we propose an IoT project for “smart environments” that monitors, generates alarms and makespredictions for air quality of the surrounding environment. Its purpose is to provide the necessary information to the user about the quality of environmental parameters, to store data for analysis using Cloud platforms, and to generate alarms. We also implement aMachine Learning algorithm to make predictionsfor the obtained parameters. Experimental evaluation shows the efficacy of the system in monitoring, storing, data analyzing and prediction.Xhulja Shahini - Distributing Translating Embeddings forModeling Multi-relational Data
Paluno - The Ruhr Institute for Software Technologyxhulja.shahini@paluno.uni-due.com
Abstract. The interest that Knowledge graph embeddings (KGE) models have re-ceived, has significantly increased over the last years. Training KGE models is a te-dious process that requires high CPU and memory capacity. Different industries aretraining their huge, often distributed datasets using KGE models. An urge to scaleKGE models has evolved, to ensure that the researcher’s hard work is being alignedwith the needs of the industry of big data. To tackle the issue of resource availabilityor usage limitations, we have come up with an approach for distributing both trainingand testing of Translating embeddings (TransE) model [1], using a cluster with sharedmemory layer.Iva Tarelli, Vanjela Pine - WOWTUNE: AUTOMATED PLAYLIST GENERATOR
Faculty of Information Technology, Polytechnic University of Tirana, Tirana, Albania {iva.tarelli, vanjela.pine}@fti.edu.al
Abstract. Nowadays, whenever we want to listen to music on our mobile phone, on the radio,pc, or any other portable device, we are most likely to refer to a playlist –a collection of several tracks -to be played in a certain sequence. Existing music players display playlists that are elaborated based on the users’ taste, but they do not feature relationships between tracks. Some of the latter introduced tools generate automated playlists relying on tracks’ metadata information, creating a predictive and narrow playlist, which can be repetitive if the same input is used. In this paper we analyze some of the existing tools for playlist generation and present WowTune, a multi-layer Automated Playlist Generator tool based on web services. We then drill down to the details of the logic this tool relies on and compare the selected tracks with the ones generated by the existing tools. The resultsof our research showthat the backend logic used by WowTune tool to generate an unpredicted playlist, succeeds in creating a pleasant playlist for the users and close to their music preferences.Session 3: Multimedia System and Services and Communication Theory
Aleksandër Biberaj, Eni Haxhiraj, Ardit Skënduli, Elson Agastra - Review of the transmission of television content via the Internet and OTT TV Live platforms
Faculty of Information Technology, Polytechnic University of Tirana, Tirana, Albania {abiberaj, eagastra}@fti.edu.al Digitalb, Technical Department Tirana, Albania{eni.haxhiraj, ardit.skenduli}@digitalb.tv
Abstract. The distribution of televisioncontent viathe Internet is an innovative technology, which is being used increasinglyfrom OTTservice providers andtraditionalbroadcast companies. Although it isan important part of the industry, there is an insufficient amount of information regarding its technology and operation. Thispaper gives a complete view of the OTT TV systems, including the technical challenges, transmission protocols and the infrastructure. This explanation is provided using a ‘demo’ OTT Live platform with one channel and severalclients.Buildingthis system, revealed that there are not a lot of difficulties in creating an OTT platform, which brings a lot of benefits to the service providers and to the users. However there are a lot of aspects that can be improved through proper research like transmission latency, advertisementinclusion and maintaininga good qualityof experience for each viewerin every device.Alban Rakipi, Aleksandër Biberaj, Indrit Enesi - SNR Performance of a Delta-Sigma Analog to Digital Converter
Faculty of Information Technology, Polytechnic University of Tirana, Tirana, Albania {arakipi, abiberaj, ienesi}@fti.edu.al
Abstract. In this paperis studied the Signal-to-Noise Ratio (SNR) performance ofa delta-sigmaAnalog to Digital converter(ADC). Analog to digital converters are crucialcomponents in almostany electronic systemsand the choice of an ADC architecture is driven by the application.The ADCs used in these systems maybe affected by differentnoise sources.Through simulations we evaluate the impact of filter length, oversampling ratio, transition bandwidth and decimation factor in SNR values of the ADC.Optimal values of these parameters are suggested in order to reduce the noise effect and consequently improve SNR.Aleksander Duka and Marenglen Biba - Radio Planning of an UMTS Operator:An Experimental Evaluation
University of New York in Tirana, Albania{aleksanderduka, marenglenbiba}@unyt.edu.al
Abstract. Mobile applications that use data, video, and voice, depend on the performance of the network. Nowadays the demand for higher capacity and quantity of data, has led to significant development of mobile networking systems. This paper deals with the construction of an UMTS network and the use of this network as efficiently as possible. The goal is a full coverage network with high capacity in the Tirana area. The work is based on prediction and calculation of the signal coverage. In the conducted experiments,the location and the type of antenna will be used for the establishment of the network. In addition, we have used a tool for predicting signal coverage. We have tried to reach full coverage by using the lowest number of antennas. In our experiments, we verified that the tool is not accurate for urban areas with greater density and that there is relevant difference between theoretical modeling and experiments in practice. We provide experimental evidence for our conclusions and we provide a critical discussion on planning processes.Elson Agastra, Orjola Jaupi - Continuous wave Doppler radar implementation working in 10.7 GHz frequency band
Faculty of Information Technology, Polytechnic University of Tirana, Tirana, Albania {eagastra, orjola.jaupi}@fti.edu.al
Abstract. This article addresses a detailed study of Doppler radars working in the 10.7 GHz frequency band to measure the speed of a moving object and distance of this object from the radar. The focus of this study is to use a real laboratory radar system which applies Doppler effect for speed and distance measurement. Radar used for this purpose is a continues wave doppler radar. Different objects with different RCS and polarization are used to prove the effectiveness of using radars which use doppler effect technique for detection, speed measurement and distance calculation of an object.Session 4: Cyber Security and Privacy
Giulia Rafaiani, Massimo Battaglioni, Marco Baldi, Franco Chiaraluce - Cyber risk assessment: a pragmatic approach
Department of Information EngineeringMarche Polytechnic University, Italy
{g.rafaiani,m.battaglioni,m.baldi,f.chiaraluce}@univpm.it
Vilma Tomço, Rexhion Qafa - Cybersecurity policy-making at Local Government: Albanian case on analysis of infrastructures, preparedness, threats and processes to reach the level of protection as central government
National Authority for Electronic Certification and Cyber Security, Tirana, Albania{vilma.tomco, rexhion.qafa }@cesk.gov.al
Abstract. Pursuant to the legislation in force for cyber security, NAECCS in Albania aims increasing the level of security in networks and information systems in the Republic of Albania. Actually, there are identified 90 CII and there is a continuous process for increasing those numbers. The results from the survey conducted by NAECCS, have shown a significant gap between central and local government related to cybersecurity, in providing the necessary infrastructure, and guarantee a certain level of cyber readiness. Based on the European directive NIS, which categorizes both the critical infrastructure and the local government, whichmanages the water, tax and other important infrastructures that directly impact the citizen, it is carried out an assessment of cyber readiness on local government. This article considers the situation of three municipalities of different size and will give recommendation on the processes to be implemented in order to reach e certain level of cybersecurity.Elior Vila, Silvester Hasani - Trusted Computing: Challenges and Opportunities for Secure Computing
University of Elbasan „Aleksander Xhuvani“, Albania elior.vila@uniel.edu.alSofia University “St. Kliment Ohridski”, Bulgariahasani@uni-sofia.bg
Abstract. Trusted Computing is an innovative technology that aims to offer security solutions by combining software and hardware protection capabilities. Today, most computing and communication systems, utilize software approaches that are supposed to be installed and configured correctly. However, such software may be affected by other software present on the same platform or by malicious software that has compromised the operating system. Trusted Computing makes use of extra cheap hardware added to the platform which provides flexible security mechanisms. The implementation of such mechanisms may present some challenges to be faced with. The main goal of this paper is to investigate such challenges and at the same time to explore the opportunities of Trusted Computing towards the security enhancement in specific areas. The focus of our work will be the Trusted Platform Module that provides the key cryptographic and protection functionalities.Marin Aranitasi, Valma Prifti, Anisa Gjini, Afërdita Mazniku - A Survey on Android Malware Detection Techniques
Faculty of Information Technology, Polytechnic University of Tirana, Tirana, Albania {marantiasi, agjini}@fti.edu.alFaculty of Mechanical Engineering, Polytechnic University of Tirana, Albaniavprifti@fim.edu.al Electronic and Postal Communication Authority (AKEP) Tirana, Albaniaaferditamazniku@gmail.com
Abstract. The number of Android user has been increasing exponentially over the past years. Linearly,has increased the risk for those users, due to the exponential growth of Android malware. A lot of effort has been done to mitigate those risks form the industry and the scientific community. This paper tries to present a review of the literature on Android security and more specifically on Android detection techniques and tools. Theaim is to present in a synthesized way the most recent discoveries that help increasing the security of theAndroidmobile devices.Olimpjon Shurdi, Ergest Alite, Aleksandër Biberaj, Genci Mesi - Designing a point to point security system with a self-defended mechanism in Cloud Computing using Snort and XGBoost algorithm
Faculty of Information Technology, Polytechnic University of Tirana, Tirana, Albania {oshurdi,ergest.alite,abiberaj, genci.mesi}@fti.edu.al
Abstract. Computing Technology is one of the Information and Communications Technologies (ICTs) which was given themost attention during this recent decade. This technology has played a key role during outbreak of pandemic situation caused by Covid-19 virus. Due tosuch fact, there is also an increased malicious activity frommalicious hackergroups which has shown their interest to benefit from possible gaps and weak points of such infrastructures and services. To identify this,we have conducted a research method implementing a point to point security system equipped with self-defended mechanism in Cloud Computing using Snort and XGBoost algorithm. This paper will present the protection of new proposed mechanism against five different types of security attacks, including Denial of Service DoS, Probing,Remote to Local Distance (R2L), User to Root attacks (U2R) and Distributed Denial of Service (DDoS) attacks.Session 5: Advanced Technology in Education and Management Systems
Enrico Zero, Alessandro Bozzi, Simone Graffione, Paolo Lambruschini, Marco Raggio - Educational case studies for system of systemsengineering
Department of Informatics, Bioengineering, Robotics and Systems Engineering, University of Genoa enrico.zero@dibris.unige.it,{alessandro.bozzi,simone.graffione}@edu.unige.it,Department of Naval, Electrical, Electronic and Telecommunications Engineering, University of Genoa paolo.lambruschini@epocs.itUniversity of Genoa marco.raggio@unige.it
Abstract. In this paper, some effective training methods are reported to explain lin-ear control systems to a class of undergraduate students during the Summer SchoolSysE 2021 in Imperia. The learning methods are composed of two parts with sometheoretical and practical lessons. The focus of the proposing approach is to use thetheoretical notions in a real case study. In this example, students test linear controlsystem theory in a homemade toy car previously assembled. Three different experi-ments using the PID controller are performed: following a specific track drawn on awhite paper, stopping as close as possible to an obstacle, using of wheels encoder totravel an exact distance. During these tests, the PID controller has been designed andtuned to automate the minicar’s displacement.Ezmerina Kotobelli, Zylal Minollari, Sanije Çela - The design of an appropriate management system for small libraries
Faculty of Information Technology, Polytechnic University of Tirana, Tirana, Albania {ekotobelli,zylal.minarolli, scela}@fti.edu.al
Abstract. Nowadays, computer programs have played an essential role in the management of libraries or bookstores.Through them we can quickly load books into the system,also with high efficiency, short operating time, low cost, and other advantages. However, there are large library management systems used to manage small or medium-sized libraries, leading to a waste of resources. Thus, software developers propose to solve the problem of managing not too large libraries by using Java as the programming language.Inthis paper we will introduce the development of a computerized management system that responds to the requirements of small libraries.Marjola Puka - Teaching experience from transmission line theory
Faculty of Electrical Engineering,Polytechnic University of Tirana,Tirana, Albaniamarjola.puka@fie.edu.al
Abstract. The study of transmission lines forms an integral part of any curriculum in electrical engineering. Studying and teaching the students the subject of electromagnetic fields for many years, I have noticed that some interesting facts about transmission linesare not available in textbooks. We have encountered during the details study in the teaching cycle with students. In this paper, I`ve treated a special problem, regarding the value of the reflection coefficient. Despite the fact that theoretically, this value is less than the unit, this paper treats the conditions and the model of a real transmission line with a reflection coefficient greater than the unit. The applied methodology and conclusion presented are fully in meeting to the power balance, the mainprinciple of circuit analysis, and transmission line theory.Enida Sheme, Evjola Spaho, Aleksandër Biberaj - Higher Education Curricula assessment based on Labor Market Survey for the ICT field in Albania
Faculty of Information Technology, Polytechnic University of Tirana, Tirana, Albania {esheme,espaho,abiberaj}@fti.edu.al
Abstract. Curricula assessment and update are essential pillars of academic activity, especially inthe rapidly changing field of Information and Communication Technologies (ICT). To be competitive and improve the employability chances of our students, we should continuously ensure the match between the curricula and the labor market demand. In this paper, we present two labor market surveys we have conducted in order to obtain a broader perspective on this issue from students’ and companies’ side. First survey was sent to Faculty of Information Technology alumni and the second one was sent to companies / institutions of ICT field in Albania hiring these alumni. We collected a total of 145alumni and 45companyrespondents. We illustrate the results of our surveys and, based on the conclusions we get from these results, we recommendfurther steps to undertake for curricula improvements at our Faculty.Session 6: Network Operations and Management
Vilma Tomço, Alma Hyra, Argena Prendi - Monitoring and Notification on Prevention and Reduction of Threats and Malwares at Network Level
National Authority for Electronic certification and cyber security, Tirana, Albania{vilma.tomco, alma.hyra,argena.prendi}@cesk.gov.al
Abstract. Malware is a constant threat for all users/computers and moreover the pandemic situation has shown an increase of such threats all over the world. It is not enough the usage of security programs: antiviruses, awareness campaign on how to navigate safe in internet, implementation of systems for activity monitoring. Considering in a wider range, monitoring network communication is still one of the primary methods that is used to combat and prevent threats. This article will describe how regular, periodic ISP network monitoring and reporting, contribute in reduction of malwares and prevention of cyber attacks, mostly at critical and important information infrastructures. Lastly, it is emphasized the need for real-time communicationin order to shorten the time of response and more effective mechanism to improve the security of end users. In the context of cleanup from threats, the role of Internet Service Providers has become more salient over time, but need to cooperate with relevant regulatory institutions.Igli Tafa, Valbona Dhjaku, Alban Rakipi, Klotilda Kola, Dajana Xhani - Interoperability of integrated systems and their efficiency in relation to traffic volume, resource management and security
Faculty of Information Technology, Polytechnic University of Tirana, Tirana, Albania {itafaj, arakipi, klotilda.kola, dajana.xhani}@fti.edu.al Natural Science Faculty, Department of Infomatics, University of Tirana, Tirana, Albaniavdhjaku@gmail.com
Abstract. In today's trade industries, how to do business online is embraced. In addition, it is extremely necessary and efficient to integrate all hardware and software parts into one. Systems integration is a major challenge in many disciplines, with a large number of technical, project, organizational or environmental problems occurring as a result of improper integration. Information technology has now become a necessity for every business and organization whether large or small. This paper shows the great importance of integrating information systems in various fields such as education (e-education), trade (e-commerce), government services (e-governmet) enabling a economic and social development of the country.Igli Tafa, Narasimha Rao Vajjhala - Exploratory Reviewof Privacyand Security Issues in the Deployment of 5G Networks
Faculty of Information Technology, Polytechnic University of Tirana, Tirana, Albaniaitafaj@fti.edu.al University of New York Tirana, Tirana, Albanianarasimharao@unyt.edu.al
Abstract. The increasing proliferation and availability of Internet of Things (IoT) devices and applications have created a growing demand for 5G networks in developing countries. Although developed countries already have several legal frameworks that they are working on to prepare for the challenges brought forward by implementing 5G networks, developing countries are way behind in designing a robust legal framework backed by ethical standards and guidelines. One of the primary issues is that of privacy. Privacy issues raise several ethical concerns, including the ownership of the data, the responsibility of the manufacturer, and regulations on data collection. This study explores the ethical issues concerns involving the implementation of 5G networks. This study develops an ethical framework that could be applied in developing countries seeking to implement 5G networks. This study explores the current best practices adopted by developed countries and identifiesaspects replicated in developing countries. 5G networks offer several prospects for startups seeking to take advantage of the increasing use of IoT devices across developed and developing countries. This study introduces various security and privacy concerns surrounding the deployment of 5G networks.Evjola Spaho, Hasan Voci - An Energy-aware Approach for a DTN RoutingProtocol: Implementation and PerformanceEvaluation
Faculty of Information Technology, Polytechnic University of Tirana, Tirana, Albania {espaho, hasan.voci}@fti.edu.al
Abstract. In this paper, an energy-aware version of Congestion Avoidance ShortestPath Routing protocol (CASPaR) is implemented and evaluated through simulations.The proposed protocol uses the threshold algorithm to determine, depending on theremaining node energy, whether a node should decide accepting or forwarding a mes-sage to its own neighbors. Different simulations are carried out with the OpportunisticNetwork Environment (ONE) simulator. Evaluation results show that the proposed andimplemented EA-CASPaR protocol reduces the number of dead nodes in the networkand the remaining energy of nodes is improved.Session 7: Intelligent Systems and Machine Learning for the Albanian Language
Kleona Binjaku, Joan Janku, Elinda Meçe - Deep learning for Albanian language identification in speech recordings
Faculty of Information Technology, Polytechnic University of Tirana, Tirana, Albania {kleona.binjaku,joan.janku,ekajo}@fti.edu.al
Abstract. The purpose of this paper is to create a system that identifies Albanian language in speech recordings as a preliminary step necessary for "speech to text" transcription. The problem is simplified in classifying the Albanian language into a set of audio signals of 2 languages: Albanian and English. The proposed system is based on generating audio signals spectrograms and using them as input to Neural Networks to make language identification. Artificial Neural Networksand Convolutional Neural Networks are used, and a comparison is made between them, to find out which model is better. The achieved accuracies are 85% (ANN) and 96% (CNN), respectively. Both, artificial neural network as well as convolutional neural networks are capable to learn spoken Albanian language patterns from their spectrogram representations.Alba Haveriku, Neki Frashëri - Building asentiment analysis modelfor tweetsin Albanian language
Faculty of Information Technology, Polytechnic University of Tirana, Tirana, Albania alba.haveriku@fti.edu.alAcademy of Sciences of Albanianeki.frasheri@gmail.com
Abstract. The rising level of social media’s usage is generating a high volume of data made available to better understand human opinions, thoughts and behavior. Noticing a small number of studies for the processing of social media data in Albanian language, the aim of this workisto presenta model to automatically process posts in Albanian, starting from the process of collecting information, cleaning and classifying, until the generation of meaningful results. The techniques presented and used are involved in the process of entities analyses,text analysis and opinion mining, providing an opportunity to better understand Albanian’s thoughts and opinions related to different subjects.We have used three machine learning classifiers (SVM, Logistic Regression, Naïve Bayes)to analyze the sentiment of Albanian tweets and evaluated their performance using statistical accuracy tests. The built model can be used to determine insights for different topics or public persons.Muhamet Kastrati, Marenglen Biba - Sentiment Analysis for Low-Resource Languages Using Deep Learning: An Overview
University of New York Tirana, Tirana, Albaniamuhamet.kastrati@gmail.commarenglenbiba@unyt.edu.al
Abstract. Sentiment analysisaims to automatically classifyend useropinions, attitudes and emotions expressed towards a particular entity, topic or event. Deep learning hasbeen widely used andhasproved usefultosentiment analysisproblemsin recent years. However, most research efforts are devoted to the rich-resource languages -there are only a few research works on sentiment analysis for the low-resource languages. Therefore, the main objective of this paper is to provide an overview of prominent deeplearning modelsthat aresuccessfully applied in sentiment analysis for low-resource languages such as Albanian language. It begins with a short introduction, motivation and some basic conceptson sentiment analysis and most prominent deep learningmodelsapplied for sentiment analysis. Next, contributions of various researchers are highlighted with the focus on deep learning approachesfollowed by some of the currenttrendsin the field.The paperconcludeswithdiscussionsand limitationsofdeep learning approach in sentiment analysis for low-resource languages.Nelda Kote, Marenglen Biba - A Neural Network Model for Opinion Mining in Albanian Language
Faculty of Information Technology, Polytechnic University of Tirana, Tirana, Albaniankote@fti.edu.alUniversity of New York Tirana, Tirana, Albaniamarenglenbiba@unyt.edu.al
Abstract. Analyzing opinionsand gaining information from opinions in online media is of interest in different aspects of everyday life. Machine learning techniques are widely used for opinion analysis. Our paper aims to evaluate the performance of the artificial neural network model in the task of text opinion classification based on the polarity of the expressed sentimentusing atwo-classes annotation schema: positive and negative. We have implemented and trained a CNNmodel using the Keras framework and Tensorflow as a backend. Theproposedmodel has an accuracy of 73%.Session 8: Image classification and object detection
Abdusamed Kura, Evis Plaku - A Convolutional Neural Network Approach forImage Classification
Canadian Institute of Technology, Tirana, Albania{abdusamed.kura, evis.plaku}@cit.edu.al
Abstract. The ability of a computer system to recognize objects quickly and effi-ciently is an important task. This paper presents the development of a machine learn-ing framework used to perform the tasks of image recognition and classification rely-ing on the implementation of a Convolutional Neural Network. Multiple layers of thenetwork are cast upon datasets of image data with the purpose of implicitly extractingfeatures and understanding patterns that are used for classification. The neural networkis trained on large sets of image data with the objective of learning how to correctlyclassify previously unseen images. A series of experiments is performed to validatethe effectiveness of the proposed approach, on two separate datasets comprising thou-sands of images of fruits and traffic signs. Experimental results show that the proposedclassification framework distinguishes objects of such disparate categories with highaccuracy.Ledio Zajmi, Arbi Bamllari - Real Time-Cloud BasedAIObject DetectioninUAS
SKAITECH, Tirana, Albania{ledio.zajmi, arbi.bamllari}@skaitech.al
Abstract. While using drones wehave the possibility to extract essential data for different industrysectorssuch as traffic analysis, people counting, license plate detection, powerlineinspectionetc. Many of the solutions in the market useeitherembedded AImodules and manufacture drones with such capabilities, or analyze the video feed via post-processing.Thisis whysuch solutions cost a lot more than the standardconsumerdroneswhich are very popular.In this paper we will present and implement ahardware agnostic cloud-basedsolution for real-time AI object detection, that can be used by consumer and commercial drones such asleadingdrone manufacturer DJI: Mavic Series(Mini, Air 2, Mavic 2 Pro,Mavic 2 Enterprise), Phantom and Matrice series.Indrit Enesi, Alban Rakipi, Desar Shahu - Increasing quality in fingerprint images by implementing wavelet fusion based on regional variance
Faculty of Information Technology, Polytechnic University of Tirana, Tirana, Albania {ienesi, arakipi, dshahu}@fti.edu.al
Abstract. Biometric identification systemsusually use fingerprint images. Accurate identification of persons depends on the quality of the fingerprint image. Usually, two or more fingerprint images are captured from the person in different positions and the fusion approach is performed. In this paper is implemented wavelet fusion based on the regional variance. Simulations show that the accuracy of minutiae detectedis increased.Organized by:
![]() |
![]() |
Faculty of Information Technology |
Polytechnic University of Tirana |
Supported by:
![]() |
![]() |
|
![]() |
![]() |
|