Discover our Team's recent Publications.

This work utilizes data from Twitter to mine association rules and extract knowledge about public attitudes regarding worldwide crises. It exploits the COVID-19 pandemic as a use case, and analyses tweets gathered between February and August 2020. The proposed methodology comprises topic extraction and visualization techniques, such as wordclouds, to form clusters or themes of opinions. It then uses Association Rule Mining (ARM) to discover frequent wordsets and generate rules that infer to user attitudes. The goal is to utilize ARM as a postprocessing technique to enhance the output of any topic extraction method. Therefore, only strong wordsets are stored after discarding trivia ones. We also employ frequent wordset identification to reduce the number of extracted topics. Our findings showcase that 50 initially retrieved topics are narrowed down to just 4, when combining Latent Dirichlet Allocation with ARM. Our methodology facilitates producing more accurate and generalizable results, whilst exposing implications regarding social media user attitudes.

Citation: P. Koukaras, C. Tjortjis and D. Rousidis, “Mining Association Rules from COVID-19 Related Twitter Data to Discover Word Patterns, Topics and Inferences”, Information Systems, p. 102054, Elsevier, 2022, Scimago Q2

A way to improve energy management is to perform balancing both at the Peer-to-peer (P2P) level and then
at the Virtual Microgrid-to-Virtual Microgrid (VMG2VMG) level, while considering the intermittency of
available Renewable Energy Source (RES). This paper proposes an interdisciplinary analytics-based
approach for the formation of VMGs addressing energy balancing. Our approach incorporates Computer
Science methods to address an Energy sector problem, utilizing data preprocessing techniques and Machine
Learning concepts. It features P2P balancing, where each peer is a prosumer perceived as an individual
entity, and Virtual Microgrids (VMGs) as clusters of peers. We conducted several simulations utilizing
clustering and binning algorithms for preprocessing energy data. Our approach offers options for generating
VMGs of prosumers, prior to using a customized Exhaustive brute-force Balancing Algorithm (EBA). EBA
performs balancing at the cluster-to-cluster level, perceived as VMG2VMG balancing. To that end, the
study simulates on data from 94 prosumers, and reports outcomes, biases, and prospects for scaling up and
expanding this work. Finally, this paper outlines potential ideal usages for the approach, either standalone
or integrated with other toolkits and technologies.

Citation: P. Koukaras, C. Tjortjis, P. Gkaidatzis, N. Bezas, D. Ioannidis, and D. Tzovaras, “A Multidisciplinary Approach on Efficient Virtual Microgrid to Virtual Microgrid Energy Balancing Incorporating Data Preprocessing Techniques”, Computing, (SpringerThis is a post-peer-review, pre-copyedit version of an article published in Computing.), Vol. 104, No. 1, pp. 209-250, 2022, Scimago Q2.

This paper presents a novel methodology using classification for day-ahead traffic prediction. It addresses the research question whether traffic state can be forecasted based on meteorological conditions, seasonality, and time intervals, as well as COVID-19 related restrictions. We propose reliable models utilizing smaller data partitions. Apart from feature selection, we incorporate new features related to movement restrictions due to COVID-19, forming a novel data model. Our methodology explores the desired training subset. Results showed that various models can be developed, with varying levels of success. The best outcome was achieved when factoring in all relevant features and training on a proposed subset. Accuracy improved significantly compared to previously published work.

Citation: S. Liapis, K. Christantonis, V. Chazan-Pantzalis, A. Manos, D.E. Filippidou, and C. Tjortjis, “Using Classification for Traffic Prediction in Smart Cities, Featuring the Impact of COVID-19”, Integrated Computer-Aided Engineering (ICAE), ), Vol. 28, pp. 417-435, IOS Press, 2021, Scimago Q1.

Application of Machine Learning (ML) and sentiment analysis on data from microblogging services has become a common approach for stock market prediction. In this paper, we propose a methodology using sentiment analysis on Twitter and StockTwits data for Stock movement prediction. The methodology was evaluated by analyzing stock movement and sentiment data. We present a case study focusing on Microsoft stock. We collected tweets from Twitter and StockTwits along with financial data extracted from Finance Yahoo. Sentiment analysis was applied on tweets, and two ML models namely SVM and Logistic Regression were implemented. Best results were achieved when using tweets from Twitter with VADER and SVM. Top F-score was 76.3% and top Area Under Curve (AUC) was 67%. SVM also achieves the greatest accuracy equal to 65.8%, when using StockTwits with TextBlob on this imbalanced data set.

Citation: C. Nousi and C. Tjortjis, “A Methodology for Stock Movement Prediction Using Sentiment Analysis on Twitter and StockTwits Data,” 2021 6th South-East Europe Design Automation, Computer Engineering, Computer Networks and Social Media Conference (SEEDA-CECNSM), 2021, pp. 1-7, doi: 10.1109/SEEDA-CECNSM53056.2021.9566242.

Easy propagation and access to information on the web has the potential to become a serious issue when it comes to disinformation. The term “fake news” describes the intentional propagation of news with the intention to mislead and harm the public and has gained more attention recently. This paper proposes a style-based Machine Learning (ML) approach, which relies on the textual information from news, such as manually extracted lexical features e.g. part of speech counts, and evaluates the performance of several ML algorithms. We identified a subset of the best performing linguistic features, using information-based metrics, which tend to agree with the literature. We also, combined Named Entity Recognition (NER) functionality with the Frequent Pattern (FP) Growth association rule algorithm to gain a deeper perspective of the named entities used in the two classes. Both methods reinforce the claim that fake and real news have limited differences in content, setting limitations to style-based methods. Results showed that convolutional neural networks resulted in the best accuracy, outperforming the rest of the algorithms.

Citation: Kasseropoulos D.P., Tjortjis C. (2021) An Approach Utilizing Linguistic Features for Fake News Detection. In: Maglogiannis I., Macintyre J., Iliadis L. (eds) Artificial Intelligence Applications and Innovations. AIAI 2021. IFIP Advances in Information and Communication Technology, vol 627. Springer, Cham. https://doi.org/10.1007/978-3-030-79150-6_51

This paper presents the development of regression and classification algorithms to predict greenhouse gas emissions caused by the building sector, and identify key building characteristics, which lead to excessive emissions. More specifically, two problems are addressed: the prediction of metric tons of CO2 emitted annually by a building, and building compliance to environmental laws according to its physical characteristics, such as energy, fuel, and water consumption. The experimental results show that energy use intensity and natural gas use are significant factors for decarbonizing the building sector.

Citation: Avramidou A., Tjortjis C. (2021) Predicting CO2 Emissions for Buildings Using Regression and Classification. In: Maglogiannis I., Macintyre J., Iliadis L. (eds) Artificial Intelligence Applications and Innovations. AIAI 2021. IFIP Advances in Information and Communication Technology, vol 627. Springer, Cham. https://doi.org/10.1007/978-3-030-79150-6_43

 

SThe effects of COVID-19 have caused severe strains to healthcare systems globally. Healthcare infrastructures are tested to their limits in almost every country and city, smart or not. This article utilizes deep and machine learning forecasting algorithms, such as Artificial Neural Networks (ANN), XGBoost and Random Forest. Using the sliding window technique, we predict the expected number of Intensive Care Unit (ICU) beds required for short (one week), mid (two weeks) and longer term (three weeks) time frames. We consider daily confirmed COVID-19 cases, current ICU, regular and special bed occupation, hospitalized cases, recovered and intubated patients and deaths. Results show that the models demonstrate very high coefficient of determination (R2) in the training phase, whilst providing accurate predictions in the forecasting phase. We report the weighted average output of ANN, XGBoost and Random Forest, which resulted in very low Mean Absolute Percentage Error (MAPE). The accurate and timely prediction of ICU beds can support decision making for Healthcare Systems, optimizing deployment of resources, as needed. Our approach can be enhanced by incorporating non-clinical parameters, based on smart city infrastructures, such as data from smart sensors.

Citation: A. Mystakidis, N. Stasinos, A. Kousis, V. Sarlis, P. Koukaras, D. Rousidis, I. Kotsiopoulos, C. Tjortjis, ‘ Predicting Covid-19 ICU needs using Deep Learning, XGBoost and Random Forest Regression with the Sliding Window technique’, IEEE Smart Cities, July 2021

 

Smart cities connect people and places using innovative technologies such as Data Mining (DM), Machine Learning (ML), big data, and the Internet of Things (IoT). This paper presents a bibliometric analysis to provide a comprehensive overview of studies associated with DM technologies used in smart cities applications. The study aims to identify the main DM techniques used in the context of smart cities and how the research field of DM for smart cities evolves over time. We adopted both qualitative and quantitative methods to explore the topic. We used the Scopus database to find relative articles published in scientific journals. This study covers 197 articles published over the period from 2013 to 2021. For the bibliometric analysis, we used the Biliometrix library, developed in R. Our findings show that there is a wide range of DM technologies used in every layer of a smart city project. Several ML algorithms, supervised or unsupervised, are adopted for operating the instrumentation, middleware, and application layer. The bibliometric analysis shows that DM for smart cities is a fast-growing scientific field. Scientists from all over the world show a great interest in researching and collaborating on this interdisciplinary scientific field.

Citation: A. Kousis and C. Tjortjis, ‘ Data Mining Algorithms for Smart Cities: A Bibliometric Analysis’, Algorithms, Vol. 14, no. 8, 242, 2021, MDPI.

 

Given the recent trend in Data Science (DS) and Sports Analytics, an opportunity has arisen for utilizing Machine Learning (ML) and Data Mining (DM) techniques in sports. This paper reviews background and advanced basketball metrics used in National Basketball Association (NBA) and Euroleague games. The purpose of this paper is to benchmark existing performance analytics used in the literature for evaluating teams and players. Basketball is a sport that requires full set enumeration of parameters in order to understand the game in depth and analyze the strategy and decisions by minimizing unpredictability. This research provides valuable information for team and player performance basketball analytics to be used for better understanding of the game. Furthermore, these analytics can be used for team composition, athlete career improvement and assessing how this could be materialized for future predictions. Hence, critical analysis of these metrics are valuable tools for domain experts and decision makers to understand the strengths and weaknesses in the game, to better evaluate opponent teams, to see how to optimize performance indicators, to use them for team and player forecasting and finally to make better choices for team composition.

Citation: V. Sarlis, V. Chatziilias, C. Tjortjis, D. Mandalidis, “A Data Science Approach Analysing the Impact of Injuries on Basketball Players and Team Performance”, Information Systems, Elsevier, 2021

 

Over the past few decades, industry and academia have made great strides to improve aspects related with optimal energy management. These include better ways for efficient energy asset management, generating great opportunities for optimization of energy distribution, discomfort minimization, energy production, cost reduction and more. This paper proposes a framework for a multi-objective analysis, acting as a novel tool that offers responses for optimal energy management through a decision support system. The novelty is in the structure of the methodology, since it considers two distinct optimization problems for two actors, consumers and aggregators, with solution being able to completely or partly interact with the other one is in the form of a demand response signal exchange. The overall optimization is formulated by a bi-objective optimization problem for the consumer side, aiming at cost minimization and discomfort reduction, and a single objective optimization problem for the aggregator side aiming at cost minimization. The framework consists of three architectural layers, namely, the consumer, aggregator and decision support system
(DSS), forming a tri-layer optimization framework with multiple interacting objects, such as objective functions, variables, constants and constraints. The DSS layer is responsible for decision support by forecasting the day-ahead energy management requirements. The main purpose of this study is to achieve optimal management of energy resources, considering both aggregator and consumer preferences and goals, whilst abiding with real-world system constraints. This is conducted through detailed simulations using real data from a pilot, that is part of Terni Distribution System portfolio

Citation: P. Koukaras, P. Gkaidatzis, N. Bezas, T. Bragatto, M. Antal, F. Carere, D. Ioannidis, C. Tjortjis and D. Tzovaras, “A Tri-layer Optimization Framework for One-day Ahead Energy Scheduling based on Cost and Discomfort Minimization”, Energies, Vol. 14, no 12, 3599; MDPI, 2021

 

A way to improve energy management is to perform balancing both at the Peer-to-peer (P2P) level and then
at the Virtual Microgrid-to-Virtual Microgrid (VMG2VMG) level, while considering the intermittency of
available Renewable Energy Source (RES). This paper proposes an interdisciplinary analytics-based
approach for the formation of VMGs addressing energy balancing. Our approach incorporates Computer
Science methods to address an Energy sector problem, utilizing data preprocessing techniques and Machine
Learning concepts. It features P2P balancing, where each peer is a prosumer perceived as an individual
entity, and Virtual Microgrids (VMGs) as clusters of peers. We conducted several simulations utilizing
clustering and binning algorithms for preprocessing energy data. Our approach offers options for generating
VMGs of prosumers, prior to using a customized Exhaustive brute-force Balancing Algorithm (EBA). EBA
performs balancing at the cluster-to-cluster level, perceived as VMG2VMG balancing. To that end, the
study simulates on data from 94 prosumers, and reports outcomes, biases, and prospects for scaling up and
expanding this work. Finally, this paper outlines potential ideal usages for the approach, either standalone
or integrated with other toolkits and technologies.

Citation: S. Liapis, K. Christantonis, V. Chazan-Pantzalis, A. Manos, D.E. Filippidou, and C. Tjortjis, “Using Classification for Traffic Prediction in Smart Cities, Featuring the Impact of COVID-19”, Integrated Computer-Aided Engineering (ICAE), IOS Press, 2021

 

A way to improve energy management is to perform balancing both at the Peer-to-peer (P2P) level and then
at the Virtual Microgrid-to-Virtual Microgrid (VMG2VMG) level, while considering the intermittency of
available Renewable Energy Source (RES). This paper proposes an interdisciplinary analytics-based
approach for the formation of VMGs addressing energy balancing. Our approach incorporates Computer
Science methods to address an Energy sector problem, utilizing data preprocessing techniques and Machine
Learning concepts. It features P2P balancing, where each peer is a prosumer perceived as an individual
entity, and Virtual Microgrids (VMGs) as clusters of peers. We conducted several simulations utilizing
clustering and binning algorithms for preprocessing energy data. Our approach offers options for generating
VMGs of prosumers, prior to using a customized Exhaustive brute-force Balancing Algorithm (EBA). EBA
performs balancing at the cluster-to-cluster level, perceived as VMG2VMG balancing. To that end, the
study simulates on data from 94 prosumers, and reports outcomes, biases, and prospects for scaling up and
expanding this work. Finally, this paper outlines potential ideal usages for the approach, either standalone
or integrated with other toolkits and technologies.

Citation: P. Koukaras, C. Tjortjis, P. Gkaidatzis, N. Bezas, D. Ioannidis, and D. Tzovaras, “An Interdisciplinary Approach on Efficient Virtual Microgrid to Virtual Microgrid Energy Balancing Incorporating Data Preprocessing Techniques”, Computing, (Springer), 2021.

 

The sports industry utilizes science to improve short to long-term team and player management regarding budget, health, tactics, training, and most importantly performance. Data Science (DS) and Sports Analytics play key roles in supporting teams, players and experts to improve performance. This paper reviews the literature to identify important attributes correlated with injuries and attempts to quantify their impact on player and team performance, using analytics in the National Basketball Association (NBA) from 2010 up to 2020. It also provides an overview of Machine Learning (ML) and DS techniques and algorithms used to study injuries. Additionally, it provides information for coaches, sports and health scientists, managers and decision makers to recognize the most common injuries and investigate possible injury patterns during competitions. We identify teams and players who suffered the most, and the type of injuries requiring more attention. We found a high impact from injuries and pathologies on performance; musculoskeletal impairments are the most common ones that lead to decreased performance. Finally, we conclude that there is a weak positive relationship between performance and injuries based on a holistic multivariate model that describes player and team performance.

Citation: V. Sarlis, V. Chatziilias, C. Tjortjis, D. Mandalidis, “A Data Science Approach Analysing the Impact of Injuries on Basketball Players and Team Performance”, Information Systems, (Elsevier), 2021

Given the recent trend in Data Science (DS) and Sports Analytics, an opportunity has arisen for utilizing Machine Learning (ML) and Data Mining (DM) techniques in sports. This paper reviews background and advanced basketball metrics used in National Basketball Association (NBA) and Euroleague games. The purpose of this paper is to benchmark existing performance analytics used in the literature for evaluating teams and players. Basketball is a sport that requires full set enumeration of parameters in order to understand the game in depth and analyze the strategy and decisions by minimizing unpredictability. This research provides valuable information for team and player performance basketball analytics to be used for better understanding of the game. Furthermore, these analytics can be used for team composition, athlete career improvement and assessing how this could be materialized for future predictions. Hence, critical analysis of these metrics are valuable tools for domain experts and decision makers to understand the strengths and weaknesses in the game, to better evaluate opponent teams, to see how to optimize performance indicators, to use them for team and player forecasting and finally to make better choices for team composition.

Citation: V. Sarlis, C. Tjortjis,  ‘Sports Analytics – Evaluation of Basketball Players and Team Performance‘ , Information Systems, 2020

 

Smart cities continuously develop into highly sophisticated bionetworks, providing both smart services and ground-breaking solutions. These bionetworks consider Smart Cities as a mechanism to produce data from multiple sharing engines, creating new challenges towards the implementation of effective Smart Cities and innovative services. The purpose of this paper is to relate Data Mining techniques and Smart City projects along with a systematic literature review that distinguishes the main topics and methods applied. The survey emphasizes on various components of Smart Cities, such as data harvesting and data mining activities over city data collected. It also addresses two research questions: a) can we forecast electricity consumption and traffic load based on past data, as well as meteorological conditions? b) which attributes are more suitable for prediction and / or decision support upon energy consumption issues? Results have shown that for both cases, various models can be built based on weather data collected.

Citation: K. Christantonis, C. Tjortjis, A. Manos, D Filippidou and E. Christelis, ‘Smart Cities Data Classification for Electricity Consumption
& Traffic Prediction’ , Automatics & Software Enginery, 2020

 

Social Media (SM) have been established as multifunctional networking tools that tend to offer an increasingly wider variety of services, making it difficult to determine their core purpose and mission, therefore, their type. This paper assesses this evolution of Social Media Types (SMTs), presents, and evaluates a novel hypothesis-based data driven methodology for analyzing Social Media Platforms (SMPs) and categorizing SMTs. We review and update literature regarding the categorization of SMPs, based on their services. We develop a methodology to propose and evaluate a new taxonomy, comprising: (i) the hypothesis that the number of SMTs is smaller than what current literature suggests, (ii) observations on data regarding SM usage and (iii) experimentation using association rules and clustering algorithms. As a result, we propose three (3) SMTs, namely Social, Entertainment and Profiling networks, typically capturing emerging SMP services. Our results show that our hypothesis is validated by implementing our methodology and we discuss threats to validity.

Citation: Koukaras P., Tjortjis C., Rousidis D., ‘Social Media Types: Intoducing a Data Driven Taxonomy’ , Computing, Springer, 2020

 

Artificial Intelligence has been an active research field in Greece for over forty years, and there are more than thirty AI groups throughout the country covering almost all subareas of AI. One milestone for AI research in Greece was in 1988, when the Hellenic Artificial Intelligence Society (EETN) was founded as a non-profit, scientific organization on devoted to organizing and promo ng AI research in Greece and abroad. This article explores current lines of AI research in Greece and gives some history of Greek AI research since 1968.

Citation: Koubarakis M., Vouros G., Chalkiadakis G., Plagianakos V., Tjortjis C., Kavallieratou E., Vrakas D., Mavridis N., Petasis G., Blekas K., Krithara A., ‘AI in Greece: The Case of Research on Linked Geospatial Data’ , the AI magazine, AAAI, 2018

 

Molecular dynamics simulations provide a sample of a molecule’s conformational space. Experiments on the μs time scale, resulting in large amounts of data, are nowadays routine. Data mining techniques such as classification provide a way to analyse such data. In this work, we evaluate and compare several classification algorithms using three data sets which resulted from computer simulations, of a potential enzyme mimetic biomolecule. We evaluated 65 classifiers available in the well-known data mining toolkit Weka, using ‘classification’ errors to assess algorithmic performance. Results suggest that: (i) ‘meta’ classifiers perform better than the other groups, when applied to molecular dynamics data sets; (ii) Random Forest and Rotation Forest are the best classifiers for all three data sets; and (iii) classification via clustering yields the highest classification error. Our findings are consistent with bibliographic evidence, suggesting a ‘roadmap’ for dealing with such data.

Citation: Tatsis V.A., Tjortjis C., Tzirakis P., ‘Evaluating data mining algorithms using molecular dynamics trajectories’ , Int’l Journal of Data Mining and Bioinformatics, Inderscience, 2013

 

This work proposes a methodology for source code quality and static behaviour evaluation of a software system, based on the standard ISO/IEC-9126. It uses elements automatically derived from source code enhanced with expert knowledge in the form of quality characteristic rankings, allowing software engineers to assign weights to source code attributes. It is flexible in terms of the set of metrics and source code attributes employed, even in terms of the ISO/IEC-9126 characteristics to be assessed. We applied the methodology to two case studies, involving five open source and one proprietary system. Results demonstrated that the methodology can capture software quality trends and express expert perceptions concerning system quality in a quantitative and systematic manner.

Citation: Kanellopoulos Y., Antonellis P., Antoniou D., Makris C., Theodoridis E., Tjortjis C., Tsirakis N.,‘Code Quality Evaluation methology using the ISOIEC 9126 Standard ‘, Int’l Journal of Software Engineering & Applications, AIRCC, 2010

 

The massive growth of Big Data kickstarted a new era for data analytics and knowledge discovery. Data mining algorithms are employed to analyze different types of data, which reside in complex information networks. Researchers focus on producing usable knowledge by taking advantage of opportunities in various domains (e.g., healthcare, social media, energy etc.). Epidemics and disease outbreaks raised concerns about effective infectious disease management in communities around the world. Therefore, they encourage the use of AI methods for management and prevention, in order to mitigate disease spread, and contain outbreaks. This work engages in predictive analytics, utilizing classification, as well as descriptive analytics utilizing association rule mining and clustering, which are widely used in healthcare and medicine, either for predicting outbreaks or for extracting usable information from healthcare and medical data. Certain steps need to be considered when attempting to perform data analysis, such as data extraction, cleaning, preprocessing, transformation, interpretation and evaluation. The experimental part of this chapter integrates widely used datasets retrieved from the UCI Machine Learning Repository related with the healthcare domain. This chapter offers a literature review on data mining in epidemics, while thoroughly discussing all the aforementioned concepts. It also presents a complete process/cycle of the required steps to analyze data retrieved from healthcare and medical sources. Hence, the research questions addressed can be summarized to the following: Q1. Which are the pervasive types of analytics involving the domains of medicine and healthcare? Q2. How is data mining performed in the fields of healthcare and medicine? Q3. Which are the widespread techniques and methods utilized? These questions are discussed and elaborated, through a concise, informative and educational narration.

 Social media (SM) is establishing a new era of tools with multi-usage capabilities. Governments, businesses, organizations, as well as individuals are engaging in, implementing their promotions, sharing opinions and propagating decisions on SM. We need filters, validators and a way of weighting expressed opinions in order to regulate this continuous data stream. This chapter presents trends and attempts by the research community regarding: (a) the influence of SM on attitudes towards a specific domain, related to public health and safety (e.g. diseases, vaccines, mental health), (b) frameworks and tools for monitoring their evolution and (c) techniques for suggesting useful interventions for nudging public sentiment towards best practices. Based on the state of the art, we discuss and assess whether SM can be used as means of prejudice or esteem regarding online opinions on health care. We group the state of the art in the following categories: virus–illness outbreaks, anti-vaccination, mental health, social trends and food and environment. Furthermore, we give more weight to virus–illness outbreaks and the anti-vaccination issues/trends in order to examine disease outbreak prevention methodologies and vaccination/anti-vaccination incentives, whilst discussing their performance. The goal is to consolidate the state of the art and give well-supported directions for future work. To sum up, this chapter discusses the aforementioned concepts and related biases, elaborating on forecasting and prevention attempts using SM data.

Citation:  Koukaras P., Rousidis D., Tjortjis C., ‘Forecasting and Prevention mechanisms using Social Media in Healthcare’ , Advanced Computational Intelligence in Healthcare, 2020

 

This paper presents research on Information Network (IN) modeling
using graph mining. The theoretical background along with a review of relevant
literature is showcased, pertaining the concepts of IN model types, network schemas and graph measures. Ongoing research involves experimentation and evaluation on bipartite and star network schemas, generating test subjects using Social
Media, Energy or Healthcare data. Our contribution is showcased by two proof of-concept simulations we plan to extend.

Citation: D. Rousidis, P. Koukaras and C. Tjortjis, ‘Examination of NoSQL Transition and Data Mining capabilities’, 14th Int’l Conf. Metadata and Semantics Research (MTSR2020), Communications in Computer & Information Science (CCIS), Springer 2020

 

Information Networks (INs) are abstract representations of realworld interactions among different entities. This paper focuses on a special
type of Information Networks, namely Heterogeneous Information
Networks (HINs). First, it presents a concise review of the recent work on
this field. Then, it proposes a novel method for querying such networks,
using a bi-functional machine learning algorithm for clustering and ranking.
It performs and elaborates on supervised and unsupervised, proof-ofconcept modelling experiments on multi-typed, interconnected data, while
retaining their semantic importance. The results show that this method
yields promising results and can be extended and utilized, using larger, realworld datasets.

Citation: P. Koukaras, C. Berberidis, and C. Tjortjis, ‘A Semi-supervised Learning Approach for Complex Information Networks’, 3rd Int’l Conf. Intelligent Data Communication Technologies and Internet of Things (ICICI 2020) 2020

 

An estimated 2.5 quintillion bytes of data are created every day. This
data explosion, along with new datatypes, objects, and the wide usage of social
media networks, with an estimated 3.8 billion users worldwide, make the exploitation and manipulation of data by relational databases, cumbersome and problematic. NoSQL databases introduce new capabilities aiming at improving the
functionalities offered by traditional SQL DBMS. This paper elaborates on ongoing research regarding NoSQL, focusing on the background behind their development, their basic characteristics, their categorization and the noticeable increase in popularity. Functional advantages and data mining capabilities that
come with the usage of graph databases are also presented. Common data mining
tasks with graphs are presented, facilitating implementation, as well as efficiency.
The aim is to highlight concepts necessary for incorporating data mining techniques and graph database functionalities, eventually proposing an analytical
framework offering a plethora of domain specific analytics. For example, a virus
outbreak analytics framework allowing health and government officials to make
appropriate decisions.

Citation: D. Rousidis, P. Koukaras and C. Tjortjis, ‘Examination of NoSQL Transition and Data Mining capabilities’, 14th Int’l Conf. Metadata and Semantics Research (MTSR2020), Communications in Computer & Information Science (CCIS), Springer 2020

 

Recommendation systems offer valuable assistance with selecting products and services. This work checks the hypothesis that taking personality into account can improve recommendation quality. Our main goal is to examine the role of personality in Movie Recommender systems. We introduce the concept of combining collaborative techniques with a personality test to provide more personalized movie recommendations. Previous research attempted to incorporate personality in Recommender systems, but no actual implementation appears to have been achieved. We propose a method and developed the 50/50 recommender system, which combines the Big Five personality test with an existing movie recommender, and used it on a renowned movie dataset. Evaluation results showed that users preferred the 50/50 system 3.6% more than the state of the art method. Our findings show that personalization provides better recommendations, even though some extra user input is required upfront.

Citation:  Nalmpantis O. and Tjortjis C.,‘The 50/50 Recommender: a Method Incorporating Personality into Movie Recommender Systems’, CCIS Communications in Computer and Information Science, Springer-Verlag, 2017

 

This paper proposes an automated approach for supporting software maintenance using software metrics and data mining. We gather metric values from C# source code elements, such as projects, files, namespaces, classes, and interfaces. These elements are clustered together, based on their similarity, with regards to these metrics, in order to identify problematic, complex classes that might be error prone. We applied this approach to two open source software systems in C#. Results show that it supports identification of potentially problematic code parts, which require further examination and proactive maintenance.

Citation: Arshad S., Tjortjis C.,‘Clustering Software Metric Values Extracted from C# Code for Maintainability Assessment’, ACM Int’l Conf. Proc. Series,ACM, 2016

 

International competition and varying customer needs commonly cause small and medium furniture manufacturing enterprises to join dynamically- formed, ‘smart’ enterprise networks, established and operating using digital information technologies. In this paper, we propose a technological approach to support such enterprise networks which is primarily based on the use of software agents. First we outline the reasons motivating networking in furniture manufacturing enterprises and we briefly present core smart enterprise network concepts. Subsequently, we provide an overview of the main technologies currently used to support enterprise networks, and we make the case for utilising service-orientation and adaptive, (semi-) autonomous software components, such as software agents. Furthermore, we propose a four-tier software architectural framework based on software agents and web services, and we briefly describe the requirements, the architecture and main features of the e-Furn software system, which is based on that framework. Finally, we discuss the intelligent recommendation feature of e-Furn.

Citation: Karageorgos A., Avramouli D., Tjortjis C., Ntalos G.,‘Agent-based Digital Networking in Furniture Manufacturing Enterprises ‘, Communications in Computer and Information Science(CCIS) LNCS, Springer-Verlag, 2010

 

Common Machine Learning applications in sports
analytics relate to player injury prediction and prevention,
potential skill or market value evaluation, as well as team or
player performance prediction. This paper focuses on football. Its
scope is long–term team and player performance prediction. A
reliable prediction of the final league table for certain leagues is
presented, using past data and advanced statistics. Other
predictions for team performance included refer to whether a
team is going to have a better season than the last one.
Furthermore, we approach detection and recording of personal
skills and statistical categories that separate an excellent from an
average central defender. Experimental results range between
encouraging to remarkable, especially given that predictions were
based on data available at the beginning of the season.

Citation: V. Chazan–Pantzalis, C. Tjortjis, ‘ Sports Analytics for Football League Table and Player Performance Prediction’, Proc. 11th IEEE Int’l Conf. on Information, Intelligence, Systems and Applications (IISA 20) 2020.

This paper provides an analysis and proposes a
methodology for predicting traffic congestion. Several machine
learning algorithms and approaches are compared to select the
most appropriate one. The methodology was implemented using
Data Mining and Big Data techniques along with Python, SQL,
and GIS technologies and was tested on data originating from one
of the most problematic, regarding traffic congestion, streets in
Thessaloniki, the 2nd most populated city in Greece. Evaluation and
results have shown that data quality and size were the most critical
factors towards algorithmic accuracy. Result comparison showed
that Decision Trees were more accurate than Logistic Regression

Citation: A. Mystakidis, C. Tjortjis, ‘Big Data Mining for Smart Cities: Predicting Traffic Congestion using Classification’, Proc.11th IEEE Int’l Conf. on Information, Intelligence, Systems and Applications (IISA 20) 2020

 

This work proposes an efficient Content-Based (CB) product recommendation methodology that promotes diversity. A heuristic CB approach incorporating feature weighting and Locality-Sensitive Hashing (LSH) is used, along with the TF-IDF method and functionality of tuning the importance of product features to adjust its logic to the needs of various e-commerce sites. The problem of efficiently producing recommendations, without compromising similarity, is addressed by approximating product similarities via the LSH technique. The methodology is evaluated on two sets with real e-commerce data. The evaluation of the proposed methodology shows that the produced recommendations can help customers to continue browsing a site by providing them with the necessary “next step”. Finally, it is demonstrated that the methodology incorporates recommendation diversity which can be adjusted by tuning the appropriate feature weights.

Citation: D. Beleveslis, C. Tjortjis, ‘Promoting Diversity in Content Based Recommendation using Feature Weighting and LSH’ , 16th Int’l Conf. on Artificial Intelligence Applications and Innovations, 2020

 

With the advent of social media, concepts such as forecasting and now casting became part of the public debate. Past successes include predicting election results, stock prices and forecasting events or behaviors. This work aims at using Twitter data, related to songs and artists that appeared on the top 10 of the Billboard Hot 100 charts, performing sentiment analysis on the collected tweets, to predict the charts in the future. Our goal was to investigate the relation between the number of mentions of a song and its artist, as well as the semantic orientation of the relevant posts and its performance on the subsequent chart. The problem was approached via regression analysis, which estimated the difference between the actual and predicted positions and moderated results. We also focused on forecasting chart ranges, namely the top 5, 10 and 20. Given the accuracy and F-score achieved compared to previous research, our findings are deemed satisfactory, especially in predicting the top 20.

Citation: E. Tsiara, C. Tjortjis, ‘Using Twitter to Predict Chart Position for Songs’ , 16th Int’l Conf. on Artificial Intelligence Applications and Innovations, 2020

 

Over 1.25 million people are killed, and 20-50 million people are seriously injured by traffic accidents every year globally, according to the World Bank. This paper aims to identify patterns in traffic accident data, collected by Cyprus Police between 2007 and 2014. The dataset that was used includes information regarding 3 groups of accident properties: human, vehicle and general environmental or infrastructural information. Data mining techniques were used, and several patterns were identified. Five classifiers were evaluated using a preprocessed dataset, to extract accident patterns. Preliminary results indicate some of the main issues with regards to accident causalities in Cyprus that could be used for real time accident warnings.

Citation: Tasios D., Tjortjis C., Gregoriades A., ‘Mining Traffic Accident Data for Hazard Causality Analysis’ ,  4th IEEE SE Europe Design Automation, Computer Engineering, Computer Networks, and Social Media Conf., SEEDA-CECNSM, 2019

 

Sports Analytics is an emerging research area with several applications in a variety of fields. These could be, for example, the prediction of an athlete’s or a team’s performance, the estimation of an athlete’s talent and market value, as well as the prediction of a possible injury. Teams and coaches are increasingly willing to embed such “tools” in their training, in order to improve their tactics. This paper reviews the literature on Sports Analytics and proposes a new approach for prediction. We conducted experiments using suitable algorithms mainly on football related data, in order to predict a player’s position in the field. We also accumulated data from past years, to estimate a player’s goal scoring performance in the next season, as well as the number of a player’s shots during each match, known to be correlated with goal scoring probability. Results are very promising, showcasing high accuracy, particularly as the predicted number of goals was very close to the actual one.

Citation: Apostolou K., Tjortjis C., ‘Sports Analytics algorithms for performance prediction’ , IEEE 10th Int’l Conf. on Information, Intelligence, Systems and Applications, pp 67-73, 2019

 

This paper presents work on using data extracted from Twitter to predict the outcome of the latest USA presidential elections on 8th of November 2016 in three key states: Florida, Ohio and N. Carolina, focusing on the two dominant candidates: Donald J. Trump and Hillary Clinton. Our method comprises two steps: pre-processing and analysis and it succeeded in capturing negative and positive sentiment towards these candidates, and predicted the winner in these States, who eventually won the presidency, when other similar attempts in the literature have failed. We discuss the strengths and weaknesses of our method proposing directions for further work.

Citation: L. Oikonomou and C. Tjortjis, ‘A Method for Predicting the Winner of the USA Presidential Elections using Data extracted from Twitter’ , 3rd IEEE SE Europe Design Automation, Computer Engineering, Computer Networks, and Social Media Conf (IEEE SEEDA-CECNSM18), 2018

 

Many algorithms have been proposed for Association Rules Mining (ARM), like Apriori. However, such algorithms often have a downside for real word use: they rely on users to set two parameters manually, namely minimum Support and Confidence. In this paper, we propose Association Rules Mining by improving the Imperialism Competitive Algorithm (ARMICA), a novel ARM method, based on the heuristic Imperialism Competitive Algorithm (ICA), for finding frequent itemsets and extracting rules from datasets, whilst setting support automatically. Its structure allows for producing only the strongest and most frequent rules, in contrast to many ARM algorithms, thus alleviating the need to define minimum support and confidence. Experimental results indicate that ARMICA generates accurate rules faster than Apriori.

Citation: Ghafari S.M. and Tjortjis C., ‘Association Rules Mining by Improving the Imperialism Competitive Algorithm (ARMICA)’ , IFIP AICT Proc. 12th Int’l Conf. on Artificial Intelligence Applications and Innovations (AIAI 2016). Vol. 475, pp 242-254, Springer, 2016

 

International competition and dynamically changing customer demands lead SME’s to join dynamically formed, `smart’ enterprise networks aiming to increase their competitiveness and market share. Supporting such networks with decision making related to collaborations and providing adaptive user interfaces are key challenges. In this paper, we use furniture manufacturing SME’s as a case study and we provide an overview of our ongoing work on e-Furn, an agent-based system for supporting `smart’ collaboration in enterprise networks. We outline two main features of the proposed approach: (a) assisting users in typical collaboration decisions, such as product bundling and task outsourcing, and (b) providing users with dynamically adaptive user interfaces.

Citation: Karageorgos A., Avramouli D., Vasilopoulou K., Tjortjis C., Ntalos G., ‘Towards Agent-Based ‘Smart’ Collaboration in Enterprise Networks’ , 8th  Int’l Workshop on Agent-based Computing for Enterprise Collaboration (ACEC) at WETICE 2010, pp. 35-40, 2010

 

Software is playing a crucial role in modern societies. Not only do people rely on it for their daily operations or business, but for their lives as well. For this reason correct and consistent behavior of software systems is a fundamental part of end user expectations. Additionally, businesses require cost- effective production, maintenance, and operation of their systems. Thus, the demand for software quality is increasing and is setting it as a differentiator for the success or failure of a software product. In fact, high quality software is becoming not just a competitive advantage but a necessary factor for companies to be successful. The main question that arises now is how quality is measured. What, where and when we assess and assure quality, are still open issues. Many views have been expressed about software quality attributes, including maintainability, evolvability, portability, robustness, reliability, usability, and efficiency. These have been formulated in standards such as ISO/IEC-9126 and CMMI. However, the debate about quality and maintainability between software producers, vendors and users is ongoing, while organizations need the ability to evaluate from multiple angles the software systems that they use or develop. So, is “Software quality in the eye of the beholder”? This workshop session aims at feeding into this debate by establishing what the state of the practice and the way forward is.

Citation: Tjortjis C., and Visser J., Ntalos G., ‘3rd International Workshop on Software Quality and Maintainability’ , Proc. IEEE 13th European Conf. Software Maintenance and Reengineering (CSMR 2009), IEEE Comp. Soc. Press, pp. 271-272, 2009

 

This paper elaborates on how to use clustering for the evaluation of a software system’s maintainability according to the ISO/IEC-9126 quality standard. More specifically it proposes a methodology that combines clustering and multicriteria decision aid techniques for knowledge acquisition by integrating groups of data from source code with the expertise of a software system’s evaluators. A process for the extraction of elements from source code and Analytical Hierarchical Processing for assigning weights to these data are provided; k-Attractors clustering algorithm is then applied on these data, in order to produce system overviews and deductions. The methodology is evaluated on Apache Geronimo, a large Open Source Application Server; results are discussed and conclusions are presented together with directions for future work.

Citation: Antonellis P., Antoniou D.,  Kanellopoulos Y., Makris C., Theodoridis E., Tjortjis C., Tsirakis N., ‘Employing Clustering for Assisting Source Code Maintainability Evaluation according to ISO/IEC-9126’ , Artificial Intelligence Techniques in Software Engineering Workshop (AISEW 2008) in ECAI08, 2008

 

This paper presents ongoing work on using data mining clustering to support the evaluation of software systems’ maintainability. As input for our analysis we employ software measurement data extracted from Java source code.

Citation: Antonellis P., Antoniou D., Kanellopoulos Y., Makris C., Theodoridis E., Tjortjis C., Tsirakis N., ‘Monitoring the Evolution of Software Systems Maintainability ‘ , Proc. special sessions in IEEE 12th European Conference on Software Maintenance and Reengineering (CSMR 2008), 2008

 

Citation: Tjortjis C., and Wang C., ‘HybridSet: An Effective Approach to Association Rule Mining’ , Proc. 22nd  European Conf. on Operational Research EURO XXII, 2007

 

This paper presents an effective Healthy Association Rule Mining (HARM) algorithm by introducing new quality measures to generate more interesting rules using itemset nutrient information. Our previous method for analyzing healthy bying patterns from quantitative attributes (or nutrient information) by interval partitions using the classical Apriori algorithm was unable to generate quality rules. This was partly because using basic itemset support is not an appropriate approach as it only gives the total support of various fuzzy sets per nutrient and not the degree of support. In this paper we propose an effective and efficient new Fuzzy Healthy Association Rule Mining Algorithm (FHARM) that produces more interesting and quality rules. In this approach, edible attributes are filtered from transactional input data by projections and are then converted to Required Daily Allowance (RDA) numeric values. The average RDA database is then converted to a fuzzy database that contains normalized fuzzy attributes comprising different fuzzy sets. Analysis of nutritional information is then performance tests and interestingness measures to demonstrate the effectiveness of the approach and proposes further work on evaluating our approach with other generic fuzzy association rule algorithsm.

Citation: Khan M.S., Muyeba M., Tjortjis C., and F. Coenen, ‘An Effective Fuzzy Healthy Association Rule Mining Algorithm (FHARM)’ , 7th Annual Workshop on Computational Intelligence (UKCI 2007), 2007

 

It is known that well over 50% of replacement projects fail. Requirements gathering go someway to contributing to this statistic; if the requirements we gather for the new system do not match those of the system to be replaced then the project is bound to fail, at least in part. This paper proposes an empirical metric that assists measuring the confidence in the requirements extracted from a legacy system. This metric capitalises on five techniques for gathering requirements from legacy systems and caters for a number of different types of project. The metric can be used to estimate the likelihood of a project’s success or failure and is evaluated by two industrial case studies; conclusions are drawn from these and directions for further work are presented

Citation: Marchant J., Tjortjis C., and Turega M., ‘A Metric of Confidence in Requirements Gathered from Legacy Systems: Two Industrial Case Studies ‘, IEEE 10th European Conf. Software Maintenance and Reengineering (CSMR 2006), pp. 353-359, 2006

 

It is known that well over 50% of replacement projects fail. Requirements gathering go someway to contributing to this statistic; if the requirements we gather for the new system do not match those of the system to be replaced then the project is bound to fail, at least in part. This paper proposes an empirical metric that assists measuring the confidence in the requirements extracted from a legacy system. This metric capitalises on five techniques for gathering requirements from legacy systems and caters for a number of different types of project. The metric can be used to estimate the likelihood of a project’s success or failure and is evaluated by two industrial case studies; conclusions are drawn from these and directions for further work are presented

Citation: Marchant J., Tjortjis C., and Turega M., ‘A Metric of Confidence in Requirements Gathered from Legacy Systems: Two Industrial Case Studies ‘, IEEE 10th European Conf. Software Maintenance and Reengineering (CSMR 2006), pp. 353-359, 2006

 

With the explosive growth of the Internet and the World Wide Web, the amount of information available online is growing in an exponential manner. As the amount of information online constantly increases, it is becoming increasingly difficult and resource demanding to search and locate information in an efficient manner. Information overload has become a pressing research problem since current searching mechanisms, such as conventional search engines, suffer from both lowprecision and low-recall. It is clear that a more dynamic, scalable and accurate searching methodology needs to be developed to overcome these limitations. This paper proposes a methodology consisting of an amalgamation of several research areas such as Web mining and relational database systems. We develop a proof of concept prototype which consists of an agent used to extract information from individual Web pages and a dynamic multi-level relational schema to encapsulate this information for later processing. The prototype provides users with a higher level of scalability and flexibility and can be utilized for searching the Internet and Intranets across large-scale organizations.

Citation: Denaxas S. and Tjortjis C., ‘Building a multi-level database for efficient information retrieval: A framework definition’, IASTED Int’l Conf. on Software Engineering (SE 2005), pp. 163-170, 2005

 

Citation: Wang C., and Tjortjis C., ‘A New Fast Algorithm for Mining Association Rules Using Logical Operations‘, Proc. EPSRC/IEEE/IEE PG Research Conf.  (PREP 2004), pp. 33-34, 2004

 

Program and system comprehension are vital parts of the software maintenance process. We discuss the need for both perspectives and describe two methods that may be integrated to provide a smooth transition in understanding from the system level to the program level. Results from a qualitative survey of expert industrial software maintainers, their information needs and requirements when comprehending software are initially presented. We then review existing software tools which facilitate system level and program comprehension. Two successful methods from the fields of data mining and concept assignment are discussed, each addressing some of these requirements. We also describe how these methods can be coupled to produce a broader software comprehension method which partly satisfies all the requirements. Future directions including the closer integration of the techniques are also identified.

Citation: Tjortjis C., Gold N., Layzell P.J. and Bennett K., ‘From System Comprehension to Program Comprehension’, IEEE 26th Int’l Computer Software Applications Conf. (COMPSAC 02), pp. 427-432, 2002

 

Software maintenance, just like any other software engineering activity, is being conducted in an increasingly distributed manner by teams which are often virtual. This paper critically reviews existing models for Virtual Organisations, investigates issues affecting Distributed Software Maintenance Teams (DSMT) and proposes a model for selecting the appropriate Computer Supported Cooperative Work (CSCW) and Groupware tools and technologies in order to facilitate communication and resource allocation for DSMT. This model builds on current theories, classifications and major concepts in the area of CSCW and advances the way DSMT are perceived. This theoretical model is yet to be empirically evaluated and enriched so that it includes Workflow management systems.

Citation: Tjortjis C., Dafoulas G., Layzell P.J., Macaulay L., ‘A Model for Selecting CSCW Technologies for Distributed Software Maintenance Teams in Virtual Organisations ‘, IEEE 26th Int’l Computer Software Applications Conf. (COMPSAC 02), pp.1104-1108, 2002

 

The paper investigates the applicability of data mining in software reliability assessment and maintenance. The proposed methodology comprises three steps. First the input models are defined by selecting parts of the source code, such as functions, routines and variables, to populate a database. Then Clustering is applied to identify sub-sets of source code that are grouped together according to custom-made similarity metrics. Finally Association rules are used to establish inter-group and intra-group relationships. Experimental results show that the methodology can assess modularity, detect complexity and predict the impact of changes.

Citation: Tjortjis C. and Layzell P.J., ‘Using Data Mining to Assess Software Reliability’,  IEEE 12th Int’l Symposium on Software Reliability Engineering (ISSRE2001), pp. 221-223, 2001

 

In this paper we describe our approach to Text Mining by introducing TextMiner. We perform term and event extraction on each document to find features that are likely to have meaning in the domain, and then apply mining on the extracted features labelling each document. The system consists of two major components, the Text Analysis component and the Data Mining component. The Text Analysis component converts semi structured data such as documents into structured data stored in a database. The second component applies data mining techniques on the output of the first component. We apply our approach in the financial domain (financial documents collection) and our main targets are: a) To manage all the available information, for example classify documents in appropriate categories and b) To “mine” the data in order to “discover” useful knowledge. This work is designed to primarily support two languages, i.e. English and Greek.

Citation: Karanikas H., Tjortjis C., Theodoulidis B., ‘An Approach to Text Mining using Information Extraction’, Proc. PKDD 2000 Workshop on Knowledge Management Theory & Applications (KMTA2000), pp.165-178, 2000

 

 This paper presents ongoing work on using data mining clustering to facilitate software maintenance, program comprehension and software systems knowledge discovery. We propose a method for grouping Java code elements together, according to their similarity. The method aims at providing practical insights and guidance to maintainers through the specifics of a system, assuming they have little familiarity with it. Our method employs a preprocessing algorithm to identify the most significant syntactical and grammatical elements of Java programs and then applies hierarchical agglomerative clustering to produce correlations among extracted data. The proposed method successfully reveals similarities between classes and other code elements thus facilitating software maintenance and Java program comprehension as shown by the experimental results presented here. The paper concludes with directions for further work

Citation:  Rousidis D. and Tjortjis C., ‘Understanding Java Source Code by Using Data Mining Clustering’ , Proc. 10th Pan’c Conf. on Informatics (PCI’2005), 2005

 

Citation: Gioldasis G., Panagopoulou G. Sirmakesis S., Tjortjis C., Tsakalidis A., ‘A Complete Model of an Information System for Hospitals‘ , Proc. 3rd Nat’l Conf. of Medical Informatics, 1994

 

Software maintainers face challenges when making decisions to modify programs with little understanding of the overall source code organisation and the full impact of changes. Most software systems are structured as a number of subsystems, consisting of code that collaborates to provide the composed functionality to the program. An important aspect of program understanding is to perceive this subsystem structure. Cluster Analysis can be of use in deriving a meaningful  subsystem structure of a program from its source code. The idea is to represent a program as a number of entities, which are grouped in clusters representing subsystems, based on their similarity measured by means of related functionality or data use. Central issues for any clustering-based approach are the specification of program entities and their attributes, similarity metrics, and clustering strategy. We propose here Data Mining Code Clustering (DMCC), an approach for supporting software maintenance and program comprehension that uses input data extracted from a C/C++ program and produces its abstraction as a number of subsystems. New possibilities in the form of weighting rationales and a framework for similarity metrics based on these, as well as an analysis of these from a novel point of view is a main contribution. The approach was evaluated by implementing a tool that was used for experimentation with programs of various sizes and languages, in collaboration with experts. Results showed that the approach is useful for deriving accurate subsystem abstractions and identifying interrelationships amongst modules. Factors influencing the feasibility of this approach are identified and directions for improvements are discussed.

Citation:   Tjortjis C., ‘Data Mining Code Clustering (DMCC): An Approach Supporting Software Maintainers with Program Comprehension’ , Technical Report, School of Science & Technology, International Hellenic University, 2019