Proceedings ICITEE23
CONFERENCE PROCEEDINGS include only extended abstracts accepted for publication by the ICITEE23 program committee, which meet all the formatting requirements and are accepted by the management of the corresponding section.
Proceedings book of ICITEE 23 | ISBN 978-9928-805-28-7 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: Advancements in Artificial Intelligence in Albania and the Albanian Language
Adapting Handwritten Text Recognition to Albanian Language: Leveraging Synthetic Datasets and Pre-Trained Models Evis Trandafili and Stela Paturri
Abstract: This paper contributes to the implementation of a Handwritten Text Recognition system for Albanian language, employing an attention-based encoder-decoder architecture. To address the absence of a publicly available Albanian handwritten dataset, we used the following approach: generate a synthetic dataset using deep learning methods, streamline dataset creation and finally enabling training for Albanian Handwritten Text Recognition model. Considering the train data scarcity in Albanian, we employed pre-trained models for English handwriting recognition. The experiments show promising results, which in the future may serve as a performance benchmark for Handwritten Text Recognition systems while handling new and untested languages.
A Comparative Analysis of Machine Learning and Deep Learning Methods for Sentiment Analysis in the Albanian Language Alba Haveriku, Sidorela Bejdo, Nelda Kote and Elinda Kajo Meçe
Abstract: In our work we aim to present a comparison between different machine and deep learning algorithms for sentiment analysis in the Albanian language. We conducted experiments based on two datasets, respectively: the AlbMoRe corpus and the Multilingual Twitter Sentiment Classification corpus. The algorithms used are Support Vector Machine (SVM), Naive Bayes, Logistic Regression and Long-Short Term Memory (LSTM). Finally, we present the results achieved from testing several hyperparameters in the implementation of the LSTM architecture.
Sentiment Analysis of Social Media Posts in Informal Albanian Language Using Artificial Intelligence Techniques Nelda Kote, Hakik Paci, Eda Graceni and Alba Haveriku
Abstract: In this paper we present a sentiment analysis study of informal Albanian language social media posts using artificial intelligence techniques. We implemented three different classification models such as LSTM neural network, Naïve Bayes, and Support Vector Machine (SVM) algorithms. The experimental results indicate that the Neural Network model outperforms the other two algorithms. This research highlights the importance of linguistic diversity in natural language processing and showcases the potential of AI in analyzing less-represented languages online.
Assessing AI Maturity, the precondition building the Path to Responsible AI Implementation in Albania Erjon Curraj
Abstract: This paper provides a comprehensive analysis of Albania's Artificial Intelligence (AI) landscape, evaluating its maturity level through desk research and a series of interviews and engagement with key stakeholders. By examining governmental strategies, policy framework and industry practices and public sector initiatives, we identify the current state of AI development, adoption, and regulation. The study reveals how Albania is navigating the complex terrain of AI technology, emphasizing the nation's movement towards the establishment of a responsible AI framework. It highlights existing capabilities, infrastructural elements, and the educational ecosystem that underpin AI progression. Moreover, it delineates the challenges and opportunities that shape Albania's approach to harnessing AI's transformative potential while ensuring ethical standards, transparency, and accountability in line with the G7 agreement on guiding principles on AI and EU AI Act. The insights garnered from this research aim to inform policymakers, business leaders, and technologists in their collaborative efforts to advance AI in Albania responsibly.
Feature engineering for sentiment analysis on Albanian corpora Besjana Muraku, Nelda Kote and Elinda Kajo Meçe
Abstract: Extracting features and using them as attributes for sentiment analysis is a known practice for improving the accuracy of the machine learning models. Albanian language is a morphological rich language which poses several challenges on classification tasks and feature engineering. In this paper we investigate the features based on word frequency. We identify two available Albanian datasets for sentiment analysis and extract most 2000 frequent words. We apply three different approaches on extracting the features: bag of words, bag of words without stop words and bag of words with negation handling. We test four classification models for each dataset and achieve an accuracy of .949 for the Naïve Bayes.
An agglomerative hierarchical clustering method for text in Albanian Luela Prifti, Denisa Salillari and Najada Firza
Abstract: Nowadays, big data is available in many areas of science, so the need to summarize data sets into groups and extract information is important. Using the cluster analysis technique, we can explore such data based on their similarity. The degree of similarity in the data is quantitatively represented by the distance functions. In this paper, using Ward's method of cosine distance in a database with 100 Albanian texts, we achieved a model with a Dunn’s index value of 0.7209. This model separated the Albanian texts into 16 different clusters based on the frequency of words, with 87 percent of texts well classified by author.
Reforming Legalization Process: An AI-driven Model for Property Title Evaluation in Albania Erison Ballasheni, Vladi Koliçi, Ilir Shinko, Bexhet Kamo, Gani Demiraj
Abstract: The Property Legalization in Albania remains manual process performed by employees, even that there are a lot of tools and models for digitizing the entire process. In this short paper introduces a rapid solution to the problem, through the implementation of a Machine Learning model using Bran.js Neural Network, which aims to enhance efficiency, eliminate corruption, and ensure a more comprehensive and fair decision-making. The proposed model delivers a percentage-based result, streamlining the qualification decision for property ownership, based on previous property evaluations.
Session 2: Network Operation and Management
FMCW Radar based Monitoring of Distracted Pedestrians Gait with Machine Learning Antonio Nocera, Michela Raimondi, Gianluca Ciattaglia, Linda Senigagliesi and Ennio Gambi
Abstract: Using the smartphone while walking is a diffused behaviour among young adults and represents a danger near crossing roads or walking down the streets because it impairs their situational awareness. In this paper, we propose a privacy-preserving and continuous monitoring of people’s behaviours with the help of a frequency modulated continuous wave (FMCW) radar. Thanks to machine learning we are able to distinguish between the so called “smartphone zombies" gait and the normal gait with a median accuracy of 92.4%; when adding a third class of fast walkers the median accuracy becomes 87.6%.
Design of an optical fiber distribution system based on Fiber to the home technology, FTTH. A real application case using COMSOF and QGIS Ornela Gjoni, Ilir Shinko, Bexhet Kamo and Vladi Kolici
Abstract: During the last century, communication networks and respective infrastructures have been the main and biggest factory. The project in question proposes a technical solution for a new communication infrastructure in a certain area (Newton America) based on Fiber to the home - FTTH technology. This is expected to be the best way to overcome the problem related to the low penetration of Information and Communication Technology and high cost. Through this technical work, the goal is to achieve fiber service for a certain number of addresses, but the challenge lies in trying to realize this design in the best possible way, as economically as possible and adhering to world design standards. Respective data are prepared and entered in QGIS and after making sure that the data is accurate as in the field, it is entered into COMSOF software, which processes, calculates and optimizes all the data entered, making sure the output is well-organized and has the best optimization and low cost.
Empirical Evaluation of Local Distributed Algorithms for Task Allocation in Network of Nodes Dorian Minarolli
Abstract: In this paper, we examine the task allocation problem in a network of computational nodes, where each node is connected to a limited number of neighboring nodes. We have developed and compared five different local distributed task allocation (DTA) algorithms. Simulation results indicate that in terms of performance metrics, two algorithms, namely "local diffusion" (LD) and "local probabilistic," (LP) emerge as the top performers compared to the others.
VANET-Radar Integration: Challenges and strategies for issue mitigation Orjola Jaupi and Evjola Spaho
Abstract: In this work, we consider the integration of Vehicle Ad-Hoc Networks (VANETs) and radar technology to improve the functionality and effectiveness of Advanced Driver-Assistance Systems (ADAS). These systems gather data and make real-time decisions to assist the driver. They can be used for detection of obstacles and help vehicles take corrective actions leading to safer and more efficient driving. An extensive analysis of potential challenges that could occur during the integration process and the diverse range of strategies to mitigate these issues is carried out.
Analysis and regulation of intermodal transportation networks affected by disruptive events Lediana Islamaj, Vladi Koliçi, Simona Sacone, Cecilia Pasquale and Endrit Lugji
Abstract: This study focuses on analyzing and regulating disruptive events in transportation networks, using a two- stage intermodal model. The research aims to identify critical links by applying a demandsensitive betweenness centrality approach and evaluates travel times in disrupted scenarios. To enhance network resilience, a 95% disruption level is implemented in critical links. The study also adjusts splitting rates to mitigate disruption effects and maintain efficient flows in the network. Overall, this research provides valuable insights into transportation network management and resilience.
Session 3: Insights into Artificial Intelligence Systems: Ethics, Emotions, and Practical Applications
Legal and Ethical Aspects of Artificial Intelligence: An Evaluation Based on Bibliometric Analysis Yusuf Kaçar, Halil İbrahim Cebeci and Emrah Aydemir
Abstract: Artificial intelligence often permeates all aspects of life without recognizing any boundaries or rules and transforms society. Soft law texts addressing potential legal and ethical issues that may arise in this context have only recently begun to emerge. In the study, biblioshiny, the web interface of the bibliometrix library in the R program, was utilized for the analysis. Given that the study revolves around artificial intelligence, ethics, and law, the search terms were selected as "Artificial intelligence or machine learning" and "ethic" and "law or legal or regulation or juristic," with only English records considered. No time restrictions were imposed. A total of 2132 academic works were examined, and besides the strong relationship between the concept of ethics and privacy, the concepts of transparency, cybersecurity, governance, data protection, responsibility, and accountability also stood out. The number of publications has notably increased, particularly in the last 5-6 years, and it is evident that the field has garnered significant attention across various regions of the world. These studies are predominantly found in computer science-oriented journals. While the studies initially leaned towards technology-focused topics, in recent times, they have shifted towards ethical considerations centered on human and psychology.
Emotional Traces Behind the Videos: Analysis of YouTube Comments Okan Karakaya and Emrah Aydemir
Abstract: YouTube, a platform with billions of daily video views, where users leave comments for their opinions on videos. The analysis of these comments is actively utilized in various fields, ranging from customer feedback improvement opportunities, competitive analysis, brand reputation, product development, marketing, and sales to risk management. Comment analysis involves processing and making sense of big data, aiding in strategic decision-making. Additionally, methods such as sentiment analysis can reveal users' emotions and emotional reactions derived from the comments. Comments from 10 automobile introduction videos on the YouTube platform were collected, resulting in a total of 9,700 comments. Initially, these comments were examined based on view and like counts, followed by sentiment analysis categorizing them as positive or negative. While the comments were very high in the first months, the number of comments decreased in the following months due to the aging of the video topic. Additionally, when the comments are examined in terms of sentiment analysis, it is noteworthy that the numbers of both positive and negative comments are close to each other. However, it is seen that a small number of video comments contain more than 60% emotion. It seems that some videos have a high number of comments because they are more remarkable in terms of content or have a large number of members.
Predicting bachelor’s students’ academic performance with machine learning algorithms Aida Shasivari, Paola Shasivari and Aleksander Xhuvani
Abstract: This paper introduces a study focused on developing an AI model to predict the academic performance of first-year computer engineering students at the Polytechnic University of Tirana (UPT). The study uses criteria such as high school math and physics grades, overall average grade, Matura exam results and calculus grade first semester. The study's significance lies in its pioneering approach for Albania and its potential to inform academic decisions.
Data Quality Analytics for Customer-Level Dataset Brunela Karamani
Abstract: Data quality is paramount in any data-driven bank, especially when dealing with customer-level datasets. Accurate and reliable customer data is essential for making informed business decisions, enhancing customer experiences, and ensuring regulatory compliance. This article addresses common data quality challenges and examines the significance of data quality analytics in the context of customer-level datasets. Banks may fully utilize the potential of their client data and obtain a competitive advantage in the data-driven economy of today by putting in place strong data quality analytics procedures. In this study, we decided to use the Oracle EDQ platform to clean and standardize customer data from many sources to produce an effective analysis. For data practitioners, researchers, and other potential consumers, the research findings provide a useful case study of data quality analytics tools and a demonstration of the data quality analytics cycle.
The Role of Generative AI in 3D Modeling and Printing Efficiency Anduel Kuqi, Ambra Korra and Indrit Enesi
Abstract: Undoubtedly, technology is in constant grouth. One such innovation is precisely the combination of generative artificial intelligence (AI) with 3D modeling and 3D printing which offers novel solutions for design optimization, complex geometries and enhanced efficiency. What we will study in this paper is exactly the Generative Algorithm for 3D Printing (GAP). This innovative approach integrates real-time feedback loops, adversarial training, and metrics integration to optimize the 3D printing process. In this paper, we present the results of the comparative analysis of GAP with a simulated traditional 3D printing method, comparing the performance based on three study elements which are: layer adherence, printing speed, and structural stability.
Integrating Physics-Informed Machine Learning with Second-Order Discrete Traffic Models for Traffic State Estimation Kleona Binjaku, Elinda Kajo Meçe, Simona Sacone and Cecilia Pasquale
Abstract: This paper introduces a physics-informed machine learning (PIML) approach to traffic state estimation, integrating a second-order discrete traffic model with neural networks. By leveraging the physical principles governing traffic dynamics, our PIML model outperforms traditional methods, providing accurate and interpretable traffic state predictions, for a real data case study. This research offers a promising avenue for enhancing traffic management systems through physics-aware machine learning.
Sign Language Recognition through Machine Learning Techniques Evis Trandafili, Hakik Paci and Egisa Fusha
Abstract: Sign language facilitates communication among hearing disabled people, but at the same time serves as a communication barrier with the general community. Recent advancements in technology and Artificial Intelligence can bridge this communication gap. This paper investigates the main blocks employed in the implementation of a Sign Language Recognition system covering sign capturing approaches, sign dataset selection, preprocessing and finally the deep learning module that engages a CNN and a CNN-SVM module.
Session 4: Innovations in Optimization: from Communication Networks to Transportation and Supply Chains
Application of multiobjective evolutionary algorithms for solving the spectral and energy efficiency trade-off problem in Massive MIMO systems: A literature review Eni Haxhiraj, Elson Agastra and Bexhet Kamo
Abstract: Massive MIMO is one of the main technologies of the 5G generation. It enables the system to achieve high values of spectral and energy efficiency, but these two objectives can not be maximized at the same time. The trade-off between these two parameters is a multiobjective problem. In this paper, all the multiobjective evolutionary algorithms used for the optimization of this problem, are analyzed and compared. Each of the algorithms has its limitations and advantages. Their performance depends on the simulation parameters and also on the specific use case.
A New Energetic Reasoning for the Cumulative Scheduling Problem Kristina Kumbria, Jacques Carlier, Antoine Jouglet and Abderrahim Sahli
Abstract: This project aims to enhance algorithms for solving the Cumulative Scheduling Problem (CuSP) and its relevance to Resource-Constrained Project Scheduling Problems (RCPSP). The proposed approach introduces an extended notion of energy, associated with an execution interval, enabling improved evaluation of task contributions and feasibility tests during solution construction. The project involves characterizing useful intervals and comparing the new approach to existing methods.
An approach to optimize the railway resource planning process in a microscopic scale Estia Maliqari, Dritan Nace, Antoine Jouglet, Giuliana Barbarino and Elinda Kajo Meçe
Abstract: Timetable planning in a railway system consists of using the infrastructure in the best possible way, by determining the planning of rolling stock to guarantee the movement of trains, time slots to represent their schedules, as well as on-board agents available to cover a given schedule. This study aims to construct an optimized model for integrated planning of train slots and rolling stocks using a microscopic simulator. To clarify the methodology, we present a detailed solution scheme.
Traffic Simulation-Based Optimization of the Tirana Roundaboat: A Case Study Veranda Syla and Aleksander Biberaj
Abstract: In this study, we examined the current status of the roundabout adjacent to the Faculty of Economics in Tirana. It often experiences capacity issues, resulting in lengthy queues. The first objective of our research was to alleviate these queues and delays by augmenting the roundabout’s capacity. We created a simulation of the current state and evaluated the effects of three key improvements on capacity, queues, and delays: the relocation of a pedestrian crossing, the addition of an extra lane to two immediate entrances and exits utilizing the PTV Vissim simulator.
Optimizing Warehouse Storage: The Location Assignment Problem Ermal Belul, Dritan Naçe, Antoine Jouglet and Marwane Bouznif
Abstract: This article focuses on optimizing warehouse order preparation by addressing the Storage Location Assignment Problem. The large real-world instances that this industry deals with on an everyday basis exceed typical academic examples in literature and these pose computational challenges. Two main approaches are presented: a greedy algorithm that prioritizes high-demand products in cost-efficient locations, and a Maximum Flow Minimum Cost model, powered by the Hungarian algorithm that globally optimizes assignment. Real-world data analysis demonstrates that these methods significantly enhance warehouse efficiency by achieving good-quality storage organization.
Forecasting Product Demand for Inventory Optimization in the Petroleum Supply Chain Ina Dervishi, Rexhina Hoxha, Ilir Shinko, Roberto Sacile and Enrico Zero
Abstract: This study explores efficient inventory management in the petroleum industry through computer engineering, data analysis, and predictive modeling. Utilizing advanced forecasting algorithms and neural networks, the study predicts future demand patterns, incorporating seasonal variations for accuracy. The research has significant implications for stakeholders, enabling agile decision-making in the evolving landscape of petroleum logistics.
Session 5: Software Engineering and Applications
Online payments using paypal in e-commerce WEB applications, a practical implementation case Bruna Biçaku, Ilir Shinko, Bexhet Kamo and Vladi Koliçi
Abstract: The primary objective of this project involves the creation of an E-commerce Web Application, implemented for a Children's Toys store, which uses the PayPal payment gateway technology. This application was built using the ASP.NET framework in conjunction with Microsoft SQL Server Management Studio. It was designed to cater to both sellers, referred to as the owners of the online shop and customers, enabling seamless online transactions. Within the application, two distinct interfaces were developed: the Admin view, tailored for the online shop owner's use, facilitating product management and customer administration, and the Public view, intended for clients or users to browse and purchase products.
Building High-Performance ETL Systems Using Microservices and Distributed Messaging Ina Papadhopulli and Igli Balla
Abstract: In an ever-growing Internet that keeps producing more and more data every day, the design and build of ETL (Extract-Transform-Load) systems that can achieve high-performance data integration can get extremely difficult. From performing the connection to sources and targets to shaping this data through long chains of transformations, an ETL needs to be reliable, scalable, and maintainable. In this study, the main purpose was to develop such a system using some of the best software architecture practices and technologies, new and old, that offer the features to achieve such software, with as little compromise as possible. For this, relying on a more distributed approach by using the microservices architecture, a high throughput messaging cluster to maintain reliable communication between the services, and a good strategy on transformation processing, can result in an overall better performing and more scalable ETL.
Checking student attendance in classes using mobile devices Hakik Paci and Evis Trandafili
Abstract: Attendance in classes is very important in education. Sometimes it is required that students attend some classes and sometimes is recommended to attend the classes but always is important for teachers to have information about statistical information about statistics of the attendance of students in classes. There are different techniques to check attendance with technology like scanning a QR Code barcode at the entrance, using RFID tags, face recognition, etc. In this paper, we will present a technique to check when a student attended the class and when he left the class by using Wi-Fi signals sent out by their mobile phones even when they are not connected to the access point device.
Implementation of reflective programming in .NET Hakik Paci and Nelda Kote
Abstract: Programming languages are evolving very fast, but their principles are almost the same. There are two main concepts about the writing process of software, Closed Code and Open-Source Software. Most developers prefer to have the source code of the software developed by other developers so they can understand, read, and change the product very easily. Even when a product is an open source, the number of developers who modify the product is very low compared to the number of users who use the product. In this paper we will present an implementation of reflective programming in .NET to allow other developers to extend the capabilities of a product with their contributions on writing source code even if the project is not open source.
Analyzing the dynamics of ERP selection across diverse company profiles Ernest Shahini and Ana Ktona
Abstract: This research investigates the influential role of company characteristics in the ERP selection process. By examining variables such as sector, company size, revenue, and workforce, our study employs a comprehensive methodology to emphasize the nuanced relationship between these factors and the outcomes of ERP projects. Join us as we explore the intricacies of ERP selection strategies tailored to the unique profiles of organizations. Gain insights into optimizing the success of ERP initiatives in a diverse business landscape, with practical implications for enhancing decision-making in the everevolving digital business environment.
Session 6: Internet of Things and Engineering: From Home Automation to Sensing Technologies
Design and Implementation of a Smart Home Automation System based on remote control from Home Kit and Android Erilda Muka, Elvis Avdyli and Gerti Mecaj
Abstract: This paper introduces an innovative project focused on intelligent home device control, leveraging state-of-the-art technologies and the integration of artificial intelligence. The proposed model is positioned as one of the most cost-effective methods available for home automation, turning a regular household into a smart home. The aim of this paper is to enhance the overall living experience for residents by providing complete remote control and automation of various processes, including streamlined remote device control based on user-selected programs and reduced energy consumption.
A Comparative Analysis of Multi-Sensor Radar Integration for Human Activity Recognition Djazila Korti and Zohra Slimane
Abstract: In this study, we undertake a comparative analysis of model performance for Human Activity Recognition (HAR) when managing multi-source radar data. To this end, we carry out three experiments using a multiple-input model, concatenating data samples and fusing data samples. These experiments are conducted using three datasets containing micro-Doppler signatures of human activities. The results show that the use of a multi-input model represents the most efficient approach, achieving the highest level of accuracy at 94.65%.
Regenerative energy management-reuse algorithm in UPSsupplied systems Fatmir Basholli and Bexhet Kamo
Abstract: Loads/equipment especially in industrial applications and other different applications that use electric motors, draw current from the network while rotating, but in case of a sudden force (braking effect) they start to produce electricity themselves. This energy is sent back to the source they are fed and in such a case the regenerative energy should be managed. If this type of load is fed by a UPS (uninterruptible power supply), in the braking mode, the UPS applies the extra energy to its DC BUS through the reverse diodes of the output power transistors (IGBT), which causes the DC BUS voltage to rise. This extra energy, regenerative energy, may be used by other equipment in the network when is possible and if not (technically), to prevent the DC BUS rise, a resistor group driven/ controlled by the respective controller is placed on the DC BUS to “re-route” the energy usage and prevent the UPS damage. In this paper we provide an algorithm that may be used to control this process of using regenerative energy by network or by re-routing it in the group of resistors.
A personalized low cost solution for supervision and control of HVAC/r systems via Modbus protocol Jonadri Bundo, Genci Sharko, Denis Panxhi and Darjon Dhamo
Abstract: The demand for remote monitoring and control of HVAC/r systems has developed in lockstep with technological improvements, which have created new opportunities while also posing new obstacles. Solving these problems can generally be accomplished through a variety of techniques, each with its own set of costs. Only some segments of the market would be able to invest in remote monitoring and control systems, as well as research and development based on monitored data, under these conditions. This paper describes a low-cost software design that integrates the Modbus protocol with the most significant HVAC/r system characteristics.
Real-Time Detection of Unusual Events in the Petrol Products Logistics Anjeza Hoxha, Kleona Binjaku, Elinda Kajo Meçe, Roberto Sacile and Enrico Zero
Abstract: The modern era has witnessed advancements in technology and communication that offer potential solutions to the challenges of dangerous goods logistics. Real-time tracking, sensor technologies, and intelligent transportation systems can provide real-time monitoring and data collection, enabling early detection of anomalies and deviations from safe conditions. Integrating such technologies with advanced decision support systems can aid in providing timely guidance to drivers and logistics managers to mitigate potential risks.
Wearable sensors for cardiac desease monitoring using IoT Gledis Basha, Lorena Balliu and Elma Zanaj
Abstract: The rapid advancement in technology in a short amount of time is giving us the necesary tools to use in remote health monitoring. Based on non-invasive wearable sensors, communication and information technologies. This tools are providers of efficient and cost-effective solution that allows the elderly to remain comfortable in their home environment rather than in conventional healthcare facilities. Low-cost, real-time patient cardiovascular data compared to the expensive hospital monitoring is the challenge we want to overcome.
Session 7: Innovation Technology in Education
Automatic Literature Review based on AI (OpenAI) Adrian Besimi, Nuhi Besimi and Xhemal Zenuni
Abstract: Artificial Intelligence (AI) is becoming more and more crucial with applications on various domains. It also impacts education and research, making it a valuable tool to find better ways to review literature. This abstract outlines why AI matters and how we use OpenAI's API for better literature reviews and finding relevant research articles for specific topic. Our proposed methodology involves three steps: 1) First, we collect research articles from the IEEE API on a specific topic; 2) next, we cluster these articles based on similar keywords and sort them by number of citations, and 3) then, we use OpenAI to analyze the extracted articles and write short summaries for the top five groups of articles. The result is a Word document with all the summaries. AI's value lies in saving time and making the research process more manageable. With so many research articles available, AI helps collect and summarize them faster, so researchers can focus on analyzing and interpreting the data. The most crucial part is using OpenAI's API. It helps us create short, easy-to-understand summaries of the articles in the top five groups. These summaries make the complex research information simple to understand. In summary, AI and OpenAI can help us making the literature review process simpler and easier by saving time and making research information more accessible. AI's role in research is growing, and this method aligns with that trend, making it easier to share and understand research findings.
Ranking of Universities using Majority Judgment Genta Babasuli, Dritan Naçe, Ina Papadhopulli and Nelda Kote
Abstract: Many organizations and researchers have worked hard to create a ranking framework by employing different approaches and concentrating on different factors. Because of methodological variations resulting from the choice of indicators, weights, data collecting, and analysis, various rankings are produced for the same institution by several ranking systems. This study’s focus is more specific, discussing the majority judgment as a new method of ranking based on the median rating from different considered rankings. To draw quantitative conclusions, we must use the language of mathematics.
Penetration Rate and Application of IoT, IoE and Green Transformation in the ICT Curricula: Assessment of the Current Situation in Six Western Balkan Universities Evjola Spaho, Enida Sheme and Aleksander Biberaj
Abstract: This paper presents the outcomes of a survey conducted among the academic staff members in the field of Information and Communication Technology (ICT) from six Universities in three Western Balkan countries (Albania, Kosovo, and Montenegro) for the penetration rate and application of Internet of Things (IoT), Internet of Everything (IoE) and green transformation in the curricula of these universities. Through analysis of the collected data, we aim to identify the present situation, find the gaps and ways to improve the curricula.
Teaching Microservice Architectures: An Experience Report Klesti Hoxha
Abstract: Microservices architectures have gained widespread adoption globally leading to technical and organization implications in the software industry. Despite the considerable benefits, they face many challenges that often lead to incorrect implementations of them. In this experience report it is detailed the teaching approach taken in a graduate university course related to microservices. Workshop styled, it included a close to real-life simulation environment that showcased the various aspects of the implementation of a microservice architecture.
Session 8: Cyber Security, Privacy, and Cloud Computing
Systematic Literature Review for Security in Cloud Computing Anduena Dibra and Igli Tafa
Abstract: Cloud Computing has become a widely used word in our daily life. The technology has revolutionized many aspects, like saving, accessing, and processing data, offering different applications and services etc. The introduction of Cloud computing has reorganized how we build networks, changing the technology concepts we knew and used before. This is especially noticeable in the IT domain, where Cloud Computing offers scalability, flexibility and cost efficiency across all services and applications. Besides its many benefits, Cloud Computing brings a series of challenges that need to be addressed. To better understand and tackle the challenges associated with the widespread use of Cloud Computing we have analyzed 51 different papers across the years 2010 to 2023. The reviewed papers were focused on Cloud computing, security, and IT domain. As a result, we have defined that there are a lot of aspects that can be improved such as in the domains related to security, compliance, personal data protection, etc.
Quantum Computing: A Risk or Challenge to Modern Cryptographic Algorithms Elior Vila
Abstract: Quantum computing is rapidly growing due to the high expectations of new emerging computer systems which can carry out computational tasks much faster than today’s traditional binary machines. The next generation of computers will be capable of solving specific problems with great improvements in efficiency compared to classical computers. These capabilities may have great positive impacts on many complex applications that today rely on conventional computer systems. Apart from benefits, quantum computing may pose significant danger to the security of some cryptographic algorithms, thus putting them at risk of compromising their secure operation. Therefore, the development of such systems needs to be faced with challenges of design of postquantum cryptographic algorithms which must be resistant to the massive computation power of upcoming quantum computers. This paper investigates the current situation of cryptographic algorithms and analyses the possible impact of advancing quantum systems on modern cryptography. Finally, some directions in post-quantum cryptography devoted to protecting traditional computer systems are discussed.
Definition and implementation of the Static and Dynamic Application Security Testing Tools in web-based information systems applied to logistics Endrit Lugji, Ilir Shinko, Roberto Sacile and Enrico Zero
Abstract: Modern logistics heavily rely on web-based IT for effective supply chain management and service flow. The growing complexity of web apps increases security risks, especially in logistics. This study focuses on web application security in logistics, emphasizing the client-server approach. SAST and DAST tools are essential for robust security. SAST assesses source code for vulnerabilities early in development, while DAST scans active apps for flaws. Implementing both tools can enhance the security of web-based logistics systems.
A Glimpse into Digital Hardware Design in the Era of Large Language Models: A simple microprocessor design example Gani Demiraj, Ilir Shinko, Vladi Koliçi, Bexhet Kamo and Erison Ballasheni
Abstract: The launch of ChatGPT at the end of 2022 fueled a heated debate revolving around the age-old question of whether artificial intelligence (AI) will be able to outperform human intelligence in the future. For many, the answer is yes, and for some, that future is now. In this short paper we present a simple experiment we designed to let our students experience what machine intelligence can do in what was long considered a domain for well-skilled and experienced engineers: hardware design.
Session 9: Computer Vision and Applications
Efficiency and Performance Evaluation of Inception-V3 and Inception-ResNet-V2 in Image Classification
Ambra Korra, Anduel Kuqi and Indrit Enesi
Abstract: The Inception algorithms are part of deep convolutional neural network architecture developed by Google. These architectures are designed for computer vision implementations, such as image classification, object detection, and feature extraction. Inception models are used in supervised deep learning tasks and this paper explores the performance and efficiency of InceptionV3 and InceptionResNetV2 for image classification. Due to the explosive growth of visual data, it is important to classify this information to improve decision-making, healthcare diagnosis, environmental monitoring, and automating tasks in different industries. By examining these two algorithms, will indicate that Inception Resnet V2 achieved a higher validation accuracy of 84%, outperforming Inception V3, which attained an accuracy of 80%. In the other hand, Inception V3 shows an advantage in terms of computational efficiency, boasting lower computational costs. The choice between these algorithms should be guided by the specific use cases and priorities, as each algorithm presents distinct strengths suitable for various applications.
Capacitated Clustering Approaches in the Assignment of Service Stations to Multi-Depots
Lorena Balliu, Gledis Basha and Elma Zanaj
Abstract: Important breakthroughs in the last years have had a huge impact on the global healthcare scene. Recent developments in computer vision and Deep Learning allowed creation of models capable of analyzing radiological imaging data. In this paper, we will make use of the above to identify lung disease through lung graphs. The aim is to train and fine-tune machine-learning model. The findings show that ML has the potential to create applications that will assist healthcare personnel during routine or in the context of emergencies.
Capacitated Clustering Approaches in the Assignment of Service Stations to Multi-Depots Ronaldo Rexhmati, Kleona Binjaku, Enida Sheme, Roberto Sacile and Enrico Zero
Abstract: This paper addresses the challenge of solving Capacitated Clustering Problems (CCP) for a real case study and introduces two approaches, evaluating their performance with Mixed Integer Linear Programming (MILP). The first method is a greedy clustering algorithm that takes capacity constraints into account, while the second method involves an improved K-means algorithm dedicated to CCP applications, specifically for the allocation of service stations to multiple depots. Both of these methods offer effective solutions for addressing complex real-world supply chain and logistics problems.
Time Efficacy Analysis of Finger Vein Recognition using Local Line Binary Pattern Blerina Zanaj and Virtyt Lesha
Abstract: The article points on the execution time of finger vein recognition, considering the line binary pattern method. The methodology of the study consists on implementation of the local line binary pattern through Matlab, the use of a certain-sized database and then the localization of the execution time considering the database in question. Finally, the study analyzes the execution time based on a larger database and the obtained results are analyzed through sum of sine with 8 terms to give an interpolation on the necessary execution time of the algorithm for a broader vein dataset.