Business Information Systems

In recent years, big data systems are getting increasingly complex and require a deep domain specific knowledge. Although a multitude of reference architectures exist, it remains challenging to identify the most suitable approach for a specific use case scenario. To overcome this problem and to provide a decision support, the design science research methodology is used. By an initial literature review process and the application of the software architecture comparison analysis method, currently established big data reference architectures are identified and compared to each other. Finally, an Analytic Hierarchy Process as the main artefact is proposed, demonstrated and evaluated on a real world use-case.

With the growth of the inter-connectivity of the world, Big Graph has become a popular emerging technology. For instance, social media (Facebook, Twitter). Prominent examples of Big Graph include social networks, biological network, graph mining, big knowledge graph, big web graphs and scholarly citation networks. A Big Graph consists of millions of nodes and trillion of edges. Big Graphs are growing exponentially and requires large computing machinery. Big Graph is posing many issues such as storage, scalability, processing and many more. This paper gives a brief overview of in-memory Big Graph Systems and some key challenges. Also, sheds some light on future research agendas of in-memory systems.

As our society becomes more and more complex, legal documents are produced at an increasingly fast pace, generating datasets that show many of the characteristics that define Big Data. On the other hand, as the trend of Open Data has spread widely in the government sector nowadays, publication of legal documents in the form of Open Data is expected to yield important benefits. In this paper, we propose the modelling of Greek legal texts based on the Akoma Ntoso document model, which is a necessary step for their representation as Open Data and we describe use cases that show how these massive legal open datasets could be further exploited.

This study proposes a forecasting methodology for univariate time series (TS) using a Recommender System (RS). The RS is built from a given TS as only input data and following an item-based Collaborative Filtering approach. A set of top-N values is recommended for this TS which represent the forecasts. The idea is to emulate RS elements (the users, items and ratings triple) from the TS. Two TS obtained from Italy’s Amazon webpage were used to evaluate this methodology and very promising performance results were obtained, even the difficult environment chosen to conduct forecasting (short length and unevenly spaced TS). This performance is dependent on the similarity measure used and suffers from the same problems that other RSs (e.g., cold-start). However, this approach does not require high computational power to perform and its intuitive conception allows for being deployed with any programming language.

The number of available ontologies on the web has increased tremendously in recent years. The choice of suitable ontologies for reuse is a decision-making problem. However, there has been little use of decision-making on ontologies to date. This study applies three Multi-Criteria Decision Making (MCDM) algorithms in ontology ranking. A number of ontologies/alternatives and the complexity metrics or attributes of these ontologies are used in the decision. The experiments are carried out with 70 ontologies and the performance of the algorithms are analysed and compared. The results show that all the algorithms have successfully ranked the input ontologies based on their degree/level of complexity. Furthermore, the results portray a strong correlation between the ranking results of the three MCDM algorithms, thereby, providing more insights on the performance of MCDM algorithms in ontology ranking.

Growing amount of complexity and enterprise data creates a need for novel business process (BP) analysis methods to assess the process optimization opportunities. This paper proposes a method of BP analysis while extracting the knowledge about Decision-Making Logic (DML) in a form of taxonomy. In this taxonomy, researchers consider the routine, semi-cognitive and cognitive DML levels as functions of BP conceptual aspects of Resources, Techniques, Capacities, and Choices. Preliminary testing and evaluation of developed method using data set of entry ticket texts from the IT Helpdesk domain showed promising results in the identification and classification of the BP Decision-Making Logic.

The potential to predict the productivity and the specific electric-energy furnace consumption is very important for the economic operation and performance of a Consteel electric-arc furnace. In this work, these two variables were predicted based on specific operating parameters with the use of machine learning. Actually, three different algorithms were tested for this study: the BRF method of support vector machine (SVM), the light gradient boosting method (lightGBM), and the Keras system with TensorFlow as backend. The results appear to be good enough for production scheduling, and are presented and discussed in this work.

The accuracy of a forecast affects the financial result of a company. By the improvement of Management Accounting (MA) processes, the introduction of advanced technology and additional skills is prognosticated. Even though companies have increasingly adopted Predictive Analytics (PA), the impact on MA overall has not been investigated adequately. This study investigates this problem through a single case study of a German company. The interview results provide an overview of requirements and benefits of PA in MA. In the future, Management Accountants will be able to focus on business partnering, but require advanced statistical knowledge to fully benefit from PA.

Empirical research effort over Corporate Social Responsibility (CSR) is typically concentrated on a limited number of aspects. We focus on the whole set of CSR activities to find out if there is a structure in those. We take data on the four major dimensions of CSR: environment, social & stakeholder, labor, and governance, from the MSCI database. To find out the structure hidden under almost constant average values, we apply a modification of K-means clustering with its complementary criterion. This method leads us to discover an impressive process of change in patterns that we predict will continue in the future.

One of the essential steps in most business is to understand customers’ preferences. In a data-centric era, profile inference is more and more relaying on mining increasingly accumulated and usually anonymous (protected) data. Personalized profile (preferences) of an anonymous user can even be recovered by some data technologies. The aim of the paper is to review some commonly used information retrieval techniques in recommendation systems and introduce new trends in heterogeneous information network based and knowledge graph based approaches. Then business developers can get some insights on what kind of data to collect as well as how to store and manage them so that better decisions can be made after analyzing the data and extracting the needed information.

Following a common definition, ontologies can be seen as a formal specification of a conceptualisation. However, it cannot be expected that there will be no changes applied to them. Obviously, any application build on top of some ontology needs to adjust to the introduced alterations. For example, a mapping designated between two ontologies (also called an ontology alignment) is valid only if participating ontologies are fixed. In this paper we present a function that can indicate, whether or not, the aforementioned alignment needs updating, in order to follow modifications done to participating ontologies, and to avoid mapping them again from scratch.

In Big Data, SQL-on-Hadoop tools usually provide satisfactory performance for processing vast amounts of data, although new emerging tools may be an alternative. This paper evaluates if Apache Druid, an innovative column-oriented data store suited for online analytical processing workloads, is an alternative to some of the well-known SQL-on-Hadoop technologies and its potential in this role. In this evaluation, Druid, Hive and Presto are benchmarked with increasing data volumes. The results point Druid as a strong alternative, achieving better performance than Hive and Presto, and show the potential of integrating Hive and Druid, enhancing the potentialities of both tools.

Matching learning corresponds to the combination of ontology matching and machine learning techniques. This strategy has gained increasing attention in recent years. However, state-of-the-art approaches implementing matching learning strategies are not well-tailored to deal with imbalanced training sets. In this paper, we address the problem of the imbalanced training sets and their impacts on the performance of the matching learning in the context of aligning biomedical ontologies. Our approach is applied to local matching learning, which is a technique used to divide a large ontology matching task into a set of distinct local sub-matching tasks. A local matching task is based on a local classifier built using its balanced local training set. Thus, local classifiers discover the alignment of the local sub-matching tasks. To validate our approach, we propose an experimental study to analyze the impact of applying conventional resampling techniques on the quality of the local matching learning.

Business process modeling traditionally has not paid much attention to the interactive features considering the dynamism of the environment in which a business process is embedded. As context-awareness is accommodated in business process modeling, decisions are still considered within business processes in a traditional way. Moreover, context-aware business process modeling excessively relies on expert knowledge, due to a lack of a methodological way to guide its whole procedure. Lately, BPM (Business Process Management) is moving towards the separation of concerns paradigm by externalizing the decisions from the process flow. Most notably, the introduction of DMN (Decision Model and Notation) standard provides a solution and technique to model decisions and the process separately but consistently integrated. The DMN technique supports the ability to extract and operationalize value from data analytics since the value of data analytics lies in improving decision-making. In this paper, a DMN-based method is proposed for the separate consideration of decisions and business processes, which allows to model context into decisions as context-aware business process models for achieving business process variability. Using this method, the role of analytics in improving some part of the decision making can also be integrated in the context-aware business process modeling, which increases the potential for using big data and analytics to improve decision-making. Moreover, a formal presentation of DMN is extended with the context concept to set the theoretical foundation for the proposed DMN-based method.

Researchers in the retail domain today propose that, in particular, complex and non-financial goals such as ‘customer experience’ represent the new imperative and leading management objective in the age of Digital Retail, questioning the role of conventional financial measures such as revenue. However, there is no evidence in research showing the corresponding and necessary shift from financial measures to non-financial measures as subject of interest in recent years. This article aims to reveal the development of financial versus non-financial metrics used in retail research in the last ten years and thus highlight the transition from conventional retail into Digital Retail from a metrics perspective. A systematic literature review, conducted on the basis of 80 high quality journals, serves as the research method of choice and sheds light on the various range of metrics used in the last ten years of retail research. More importantly, the results still show a major focus on financial measures despite indicating a rising awareness of non-financial, more complex and intangible measures such as ‘customer experience’ or ‘customer satisfaction’. While this finding supports proposed shift towards non-financial measures in current retail research in one side, it also shows a lack of research focusing on non-financial objectives in retail, in comparison to financial measures.

The purpose of the study is to validate the ability of a maturity model for measuring escalation capability of IT-related security incident. First, an Escalation Maturity Model (EMM) and a tool were developed to measure the maturity of an organization to escalate IT-related security incidents. An IT tool for self-assessment was used by a representative from three organizations in the Swedish health sector to measure the organization’s ability to escalate IT-related security incident. Second, typical security incident scenarios were created. The incident managers from the different organizations were interviewed about their organization’s capabilities to deal with these scenarios. Third, a number of independent information security experts, none of whom had seen the results of EMM, ranked how the three different organizations have handled the different scenarios using a measurable scale. Finally, the results of EMM are compared against the measurable result of the interviews to establish the predictive ability of EMM. The findings of the proof of concept study shows that the outcome of EMM and the way in which an organization would handle different incidents correspond well, at least for organizations with low and medium maturity levels.

As more and more machine learning based systems are being deployed in industry, monitoring of these systems is needed to ensure they perform in the expected way. In this article we present a framework for such a monitoring system. The proposed system is designed and deployed at Mastercard. This system monitors other machine learning systems that are deployed for use in production. The monitoring system performs concept drift detection by tracking the machine learning system’s inputs and outputs independently. Anomaly detection techniques are employed in the system to provide automatic alerts. We also present results that demonstrate the value of the framework. The monitoring system framework and the results are the main contributions in this article.

This paper presents a novel method to determine the optimal Multi-layer Perceptron structure using Linear Regression. Starting from clustering the dataset used to train a neural network it is possible to define Multiple Linear Regression models to determine the architecture of a neural network. This method work unsupervised unlike other methods and more flexible with different datasets types. The proposed method adapt to the complexity of training datasets to provide the best results regardless of the size and type of dataset. Clustering algorithm used to impose a specific analysis of data used to train the network such us determining the distance measure, normalization and clustering technique suitable with the type of training dataset used.

The process of developing systematic reviews is a well established method of collecting evidence from publications, where it follows a predefined and explicit protocol design to promote rigour, transparency and repeatability. The process is manual and involves lot of time and needs expertise. The aim of this work is to build an effective framework using machine learning techniques to partially automate the process of systematic literature review by extracting required data elements of anxiety outcome measures. A framework is thus proposed that initially builds a training corpus by extracting different data elements related to anxiety outcome measures from relevant publications. The publications are retrieved from Medline, EMBASE, CINAHL, AHMED and Pyscinfo following a given set of rules defined by a research group in the United Kingdom reviewing comfort interventions in health care. Subsequently, the method trains a machine learning classifier using this training corpus to extract the desired data elements from new publications. The experiments are conducted on 48 publications containing anxiety outcome measures with an aim to automatically extract the sentences stating the mean and standard deviation of the measures of outcomes of different types of interventions to lessen anxiety. The experimental results show that the recall and precision of the proposed method using random forest classifier are respectively 100% and 83%, which indicates that the method is able to extract all required data elements.

The cost of substance use regarding lives lost, medical and psychiatric morbidity and social disruptions by far surpasses the economic costs. Alcohol abuse and dependence has been a social issue in need of addressing for centuries now. Methods exist that attempt to solve this problem by recognizing inebriation in humans. These methods include the use of blood tests, breathalyzers, urine tests, ECGs and wearables devices. Although effective, these methods are very inconvenient for the user, and the required equipment is expensive. We propose a method that provides a faster and convenient way to recognize inebriation. Our method uses Viola-Jones-based face-detection for the region of interest. The face images become input to a Convolutional Neural Network (CNN) which attempts to classify inebriation. In order to test our model’s performance against other methods, we implemented Local Binary Patterns (LBP) for feature extraction, and Support Vector Machines (SVM), Gaussian Naive Bayes (GNB) and k-Nearest Neighbor (kNN) classifiers. Our model had an accuracy rate of 84.31% and easily outperformed the other methods.

The topic of this paper is the interaction of Average Information Content (IC) and frequency of aspect-coded verbs in Linear Mixed Effect Models as predictors of the verbs’ lengths. For 30 languages in focus, it came to light that IC and frequency do not have a simultaneous, positive impact on the length of verb forms: the effect of the IC is high, when the effect of frequency is low and vice versa. This is an indication of Uniform Information Density [13–16]. Additionally, the predictors IC and frequency yield high correlations between predicted and actual verbs’ lengths.

Organizations utilize Key Performance Indicators (KPIs) to monitor whether they attain their goals. For this, software vendors offer predefined KPIs in their enterprise software. However, the predefined KPIs will not be relevant for all organizations due to the varying needs of them. Therefore, software vendors spend significant efforts on offering relevant KPIs. That relevance determination process is time-consuming and costly. We show that the relevance of KPIs may be tied to the specific properties of organizations, e.g., domain and size. In this context, we present our novel approach for the automated prediction of which KPIs are relevant for organizations. We implemented our approach and evaluated its prediction quality in an industrial setting.

With the growing number of available databases having a very large number of records, existing knowledge discovery tools need to be adapted to this shift and new tools need to be created. Genetic Programming (GP) has been proven as an efficient algorithm in particular for classification problems. Notwithstanding, GP is impaired with its computing cost that is more acute with large datasets. This paper, presents how an existing GP implementation (DEAP) can be adapted by distributing evaluations on a Spark cluster. Then, an additional sampling step is applied to fit tiny clusters. Experiments are accomplished on Higgs Boson classification with different settings. They show the benefits of using Spark as parallelization technology for GP.

This article presents a genetic algorithm (GA) to solve the picker routing problem in multiple-block warehouses in order to minimize the traveled distance. The GA uses survival, crossover, immigration, and mutation operators, and is complemented by a local search heuristic. The genetic algorithm provides average distance savings of 13.9% when compared with s-shape strategy, and distance savings of 23.3% when compared with the GA with the aisle-by-aisle policy. We concluded that the GA performs better as the number of blocks increases, and as the percentage of picking locations to visit decreases.

(Business) Information systems become more and more complex due to an increase in the volume of data, but also due to more and more interconnected elements that all need to be orchestrated to perform as a uniform system. Correctly understanding and describing (business) processes, is one of the cornerstone foundations in the creation of almost all information systems. While the systems themselves have become more complex, and the means to program them have evolved over the last decades, the means to analyze and communicate about the processes they execute have stagnated on a simplistic level from the 1960s. Over the last 15 years, there has been work done in the development of concepts and tools on the topic of subject-orientation and subject-oriented (business) process modeling and management (S-BPM) that is different from earlier, classical process description approaches. This paper analyzes and argues about the shortcomings and discrepancies of those classical approaches and argues how subject-orientation may be an improvement when employed as a means in the design and development of information systems.

For the permanent establishment and use of a RIS in universities and academic institutions, it is absolutely necessary to ensure the quality of the research information, so that the stakeholders of the science system can make an adequate and reliable basis for decision-making. However, to assess and improve data quality in RIS, it must be possible to measure them and effectively distinguish between valid and invalid research information. Because research information is very diverse and occurs in a variety of formats and contexts, it is often difficult to define what data quality is. In the context of this present paper, the data quality of RIS or rather their influence on user acceptance will be examined as well as objective quality dimensions (correctness, completeness, consistency and timeliness) to identify possible data quality deficits in RIS. Based on a quantitative survey of RIS users, a reliable and valid framework for the four relevant quality dimensions will be developed in the context of RIS to allow for the enhancement of research information driven decision support.

Smart wearables can be used in the workplace for organisations to monitor and decrease the stress levels of their employees so they can work better. Such technologies record personal data, which employees might not want to share. The GDPR makes it compulsory to get employees’ consent in such a scenario, but is seen as asking a yes/no question. We show that implementing this consent procedure is not enough to protect employees and make them adopt devices. Based on interviews, we argue that more control must be given to employees on which data is collected and why through an ongoing engagement and consent procedure. It could lead to higher technology adoption rates and data quality.

This paper proposes semantic approach to manage cyber threat intelligence (CTI). The economic rational is presented as well as functional needs. Several cases of domain standards, tools and practices are modeled as a representation of the CTI sub-domain. This work focuses on the technical and operational CTI that is common to most organizations.

The presented paper aims at proposing a research methodology for creation and approbation of a flexible methodology for development and management of innovative projects in scientific organizations (FMIPSO). For basement, the following flexible methodologies have been used: Lean startup, Agile, Scrum, Design thinking, User centricity and User innovation which all are extremely applied in ICT development. The creation of FMIPSO addresses the weak success of developed and realized innovations by scientific organizations and universities, especially relevant for multidisciplinary innovation projects that include ICT as well as other sciences. This lack of good innovation performance by scientific organizations further increases the distance and integrity of the science-business-related innovation industry. The research approach includes approbation of the FMIPSO by three interdisciplinary innovation projects from science institutions for providing proofs of its relevance and applicability.

Besides its application in the software development lifecycle, COSMIC Functional Size Measurement (FSM) is investigated as a means to measure the size of business processes (BP). This paper proposes a comprehensive COSMIC FSM-based framework to verify the conformity of the business process design and run-time models with their aligned information system (IS). It relies on the standard notations BPMN and BPEL to describe the business process and run-time models, respectively, and the component diagram to describe the IS. The paper defines formulas to apply COSMIC on these models and heuristics to verify their conformity. It illustrates the approach through a case study.

Although the Strategic Information Systems Planning (SISP) has been emerging for large firms, family businesses do not develop strategic planning and they do not support business goals using Information Systems (IS). Thus, the implemented plans are not effective, and they do not meet the objectives. The purpose of this paper is to indicate the phases which contribute to a greater extent of success in order to provide conclusions regarding the implementation of this process in SMEs. Data were collected from IS executives in Greek SMEs. Factor Analysis is performed on the detailed items of the SISP process and success constructs.

The paper presents a brief technology survey of existing tools to implement data ingestion pipelines in a classical Data Science project. Given the emergent nature of technologies and the challenges associated with any Big Data project, we propose to identify and discuss the main components of a data pipeline, from a data engineering perspective. The data pipeline is showcased with a case study from an ICT university project, where several teams of master students competed towards designing and implementing the best solution for a manufacturing data pipeline. The project proposes a research-based multidisciplinary approach to education, aiming at empowering students with a novel role in the process of learning, that of knowledge creators. Therefore, on the one hand, the paper discusses the main components of a Big Data pipeline and on the other hand it shows how these components are addressed and implemented within a concrete ICT project from education, realized in tight relation with the IT industry.

5G networking is expected to induce more changes in the era of mobile communications, since new players will appear, including the new vertical industries, while the old ones will undertake new roles. 5G ESSENCE [1] addresses the paradigms of Edge Cloud computing and Small Cell as-a-Service (SCaaS) by fuelling the drivers and removing the barriers in the Small Cell (SC) market. In this paper, we investigate current market status in 5G networks and small cells, as well as we apply the business model canvas methodology for the 5G ESSENCE approach in order to explore the 5G business environment.

Serverless computing is currently taking a momentum due to the main benefits it introduces which include zero administration and reduced operation cost for applications. However, not all application components can be made serverless in sight also of certain limitations with respect to the deployment of such components in corresponding serverless platforms. In this respect, there is currently a great need for managing hybrid applications, i.e., applications comprising both normal and serverless components. Such a need is covered in this paper through extending the Melodic platform in order to support the deployment and adaptive provisioning of hybrid, cross-cloud applications. Apart from analysing the architecture of the extended platform, we also explain what are the relevant challenges for supporting the management of serverless components and how we intend to confront them. One use case is also utilised in order to showcase the main benefits of the proposed platform.

Interpretable decision making frameworks allow us to easily endow agents with specific goals, risk tolerances, and understanding. Existing decision making systems either forgo interpretability, or pay for it with severely reduced efficiency and large memory requirements. In this paper, we outline DeepID, a neural network approximation of Influence Diagrams, that avoids both pitfalls. We demonstrate how the framework allows for the introduction of robustness in a very transparent and interpretable manner, without increasing the complexity class of the decision problem.

This paper briefly presents a reference architecture called Virtual Physical Space (ViPS). The purpose of the architecture is to be able to adapt to the development of various Cyber-Physical-Social applications. In the paper, a possible adaptation for a smart seaside city is discussed. The need for virtualization of things from the physical world in a formal way is also considered. Furthermore, the virtualization and modeling of spatial aspects through the AmbiNet formalism is demonstrated by an example.

The purpose of this paper is to validate a research model that explains performance expectancy of IoT from psychological and cognitive variables: personal innovativeness of information technology (PIIT) and social influence respectively. Data were collected from small and medium-sized enterprises (SMEs) in Chile. A confirmatory approach using PLSc was employed to validate the hypotheses. The conclusions of the study are (a) Chilean SMEs companies do not use IoT massively, (b) goodness of fit indicators allowed to validate the proposed research model successfully, (c) both constructs, social influence and personal innovativeness of information technology, explain 61% of performance expectancy of IoT.

The Internet of Things (IoT) is making significant advances, especially in smart environments, but it still suffers from security issues and vulnerabilities. Hence, the security approach seems to be inappropriate for IoT-based workspaces due to the centralised architecture used by most connected objects. However, BlockChain (BC) is a technology that has been recently used to enhance security in mainly peer-to-peer (P2P) networks. The main research goal of this study is to determine whether BC, IoT, and decentralised models can be used to secure Smart Learning Environments (SLEs). In this paper, we propose a new secure and reliable architecture for connected workspaces that eliminates the central intermediary, while preserving most of the security benefits. The system is investigated on an environment containing several connected sensors and uses the BC application, offering a hierarchical network that manages transactions and profits from the distributed trust method to enable a decentralised communication model. Finally, an evaluation is conducted to highlight the effectiveness in providing security for the novel system.

Currently, we are on the brink of a period of fundamental change for the healthcare expertise of specialists, i.e., existing know-how becomes less available to patients. One of the main reasons is an economic problem: most people and organisations cannot afford the services of first-rate professionals; and most economies are struggling to sustain their professional services, including schools, court systems, and health services. Another reason for the change in healthcare is the rapid growth of evidence-based medical knowledge, where a new research paper is published, on average, every forty-one seconds. Thus, evidence-based medicine is malfunctioning, and delayed, missed, and incorrect diagnoses occur in 10 to 20% of the time. Innovative IT technologies can solve the critical challenges in the healthcare domain. One of such technologies is smart contracts that manage and enforce contracts (business rules) without the interference of a third party. Smart contracts improve interoperability and privacy in cross-organisational processes. In this paper, we define problematic processes in healthcare and then provide a smart contract-based mapping for improving these processes. This paper proposes the way to overcome inequality in services accessibility, inefficient use of services and shortcomings in service quality, using smart contracts and blockchain technology.

In Cloud Data centers, virtual machine consolidation on minimizing energy consumed aims at reducing the number of active physical servers. Dynamic consolidation of virtual machines (VMs) and switching idle nodes off allow Cloud providers to optimize resource usage and reduce energy consumption. One aspect of dynamic VM consolidation that directly influences Quality of Service (QoS) delivered by the system is to determine the best moment to reallocate VMs from an overloaded or undeloaded host. In this article we focus on energy-efficiency of Cloud datacenter using Dynamic Virtual Machine Consolidation Algorithms by planetLab workload traces, which consists of a thousand PlanetLab VMs with large-scale simulation environments. Experiments are done in a simulated cloud environment by the CloudSim simulation tool. The obtained results show that consolidation reduces the number of migrations and the power consumption of the servers. Also application performances are improved.

As wearable technologies are gaining increased attention in construction, we present an integrated solution for their adoption in occupational health and safety (OHS). Research methods include a structured literature review of 37 articles and a year-long design science research project in a construction group. The main results are (1) the identification of new wearable solutions made available by industry 4.0 to prevent hazards, and (2) a wearable model for voluntary regulations compliance. For theory, our research identifies key application areas for integrated smart OHS in construction and highlights the importance of continuous monitoring and alerts to complement the traditional sampling techniques. For practice, we offer recommendations for managers wishing to implement continuous compliance checking and risk prevention using wearable technology. Our findings help improve health and safety audits supported by digital evidence in the sector with most risks of accidents in the European Union.