Case Studies and Education in Data Science At Universities in Slovakia∗ L’ubomír Antoni1 , Erik Bruoth1 , Ján Guniš1 , Peter Gurský1 , Stanislav Krajči1 , Ondrej Krídlo1 , Radim Navrátil2 , L’ubomír Šnajder1 , Gabriela Andrejková1 , and Dušan Šveda3 1 Institute of Computer Science, Faculty of Science, Pavol Jozef Šafárik University in Košice, Jesenná 5, 040 01 Košice 2 Institute of Mathematics, Faculty of Science, Pavol Jozef Šafárik University in Košice, Jesenná 5, 040 01 Košice 3 Department of Mathematics and Statistics, Faculty of Science, Masaryk University in Brno, Czech Republic lubomir.antoni@upjs.sk, erik.bruoth@upjs.sk, jan.gunis@upjs.sk, peter.gursky@upjs.sk, stanislav.krajci@upjs.sk, ondrej.kridlo@upjs.sk, navratil@math.muni.cz, lubomir.snajder@upjs.sk, gabriela.andrejkova@upjs.sk, dusan.sveda@upjs.sk Abstract: Data analysis solutions are already being used the available methods of their analysis, which allow them in many areas of technical, natural, humanitarian and eco- to discover new knowledge. The aim of is to extract new, nomic sciences. The information hidden within data can valid and potentially useful knowledge from available data help to solve the many pending issues within community, in various areas of academic and corporate life. Complet- enterprise or science. Turning of data into knowledge and ing objectives and knowledge itself leads to several phases wisdom is beneficial and necessary. Moreover, students in which it is necessary to distinguish the phase of data with knowledge of Data Science have the potential to be comprehension, data preparation, modeling, evaluation of highly desirable in the labor market. We present the se- results and putting results into practice [1, 2, 3, 4]. lected case studies in Data Science Area. We bring a re- Section 2 describes classification tasks in general. Sec- view of Data Science subjects that are introduced and in- tion 3 provides a basic overview of university subjects at novated at universities in Slovakia within the IT Academy the five partner universities of the IT Academy Project in Project. the area of Data Science, including classification tasks. In Section 4, we present the analysis of the educational tasks and objectives of five real teachers giving lessons in com- 1 Introduction puter science. The term "science" implies knowledge gained through sys- tematic study. In one definition, it is a systematic enter- 2 Classification tasks prise that builds and organizes knowledge in the form of testable explanations and predictions. Data science might therefore imply a focus involving data and, by extension, Data Science is an interdisciplinary area that uses scien- statistics, or the systematic study of the organization, prop- tific methods, processes, algorithms, and systems to ac- erties, and analysis of data and its role in inference, includ- quire knowledge and insights from data in various forms, ing our confidence in the inference [1]. both structured and unstructured. It introduces the concept The huge amount of data and the need for decision- of data unification, data analysis, machine learning and making imply the demand for data analysis and profes- related methods to analyze current data processes. Data sionals for data processing. This issue is highly relevant, Science uses techniques and practices from the fields of data analysis solutions are already being used in many ar- mathematics, statistics, or informatics [1-4]. eas of technical, natural, humanitarian and economic sci- In the past, computers were used to process numerical ence. Moreover, data analysis experts are already scarce, data. In recent decades, representation in the form of rela- and students with knowledge of Data Science have the po- tional data has also been promoted, because it allows the tential to be highly desirable in the labor market. In devel- relationship between the instances of objects to be exam- oping and addressing this area, it is necessary to address ined in a much wider dimension. Moreover, the knowl- the amount of available data that is regularly generated by edges found in the data helps to solve many issues related many commercial devices and scientific instruments, and to science, business or society. Permanent modification of input data into forms of knowledge and knowledge is ∗ This article was created in the framework of the National project IT therefore beneficial and necessary. The data are obviously Academy – Education for the 21st Century, which is supported by the European Social Fund and the European Regional Development Fund in in a tabular form, and we can obtain this form of data by the framework of the Operational Programme Human Resources. This recording the results of doctor’s examinations, customer work was partially supported by the Slovak Research and Development behavior and consumer habits, or recording the results of Agency under the contract No. APVV-15-0091. the students. Copyright c 2019 for this paper by its authors. Use permitted un- der Creative Commons License Attribution 4.0 International (CC BY Data classification is a two-step process, consisting of a 4.0). learning step (where a classification model is constructed) and a classification step (where the model is used to predict Case studies in data mining class labels for given data) [3]. (Pavol Jozef Šafárik University in Košice) In the first step, a model is created that describes a set The course is designed to deal with prepared case studies of data classes. The model is constructed based on the using data mining methods, which is an interdisciplinary analysis of input objects (called also examples, instances, topic and covers knowledge from a variety of disciplines, data points or samples) described by attributes. This is such as mathematics, statistics, and various specific areas the learning step (or training phase), where a classification of informatics. The course will introduce methods for au- algorithm builds the classifier by analyzing or “learning tomated analysis of large volumes of data and extraction of from” a training set made up of database tuples and their knowledge from these data. The course is aimed at solv- associated class labels. A tuple, X, is represented by an n- ing practical problems in the area of graphing using the dimensional attribute vector, X = (x1 , x2 , . . . , xn ), depict- presented methods and using appropriate software tools. ing n measurements made on the tuple from n database R-project as a computing environment, Programming in R: attributes, respectively, A1 , A2 , . . . , An . Each tuple, X, is data structures, graphics, applied mathematics, data anal- assumed to belong to a predefined class as determined by ysis. Java libraries for data-mining, their use. Testing another database attribute called the class label attribute. data mining algorithms. Dissemination of data-mining li- Because the class label of each training tuple is pro- braries. Solution of practical examples. Solutions for ap- vided, this step is also known as supervised learning (i.e., plication projects. the learning of the classifier is “supervised” in that it is told to which class each training tuple belongs). It con- Machine learning trasts with unsupervised learning (or clustering), in which (Pavol Jozef Šafárik University in Košice) the class label of each training tuple is not known, and the The course is focused on modeling and algorithmization of number or set of classes to be learned may not be known the learning processes of computer systems from data sets, in advance. The learned model can be represented in the the result of learning being objects that take into account form of classification rules, decision trees, or a mathe- input data sets. This modeling is based on the needs of matical formula. For example, for database of customer many areas of human activity (biomedical informatics, ex- credit information, classification rules can be used to iden- pert systems creation, robot construction, speech and text tify customers with excellent or fair credit ratings. In the processing, predictions of some events, etc.). The basis for second step, the model is used for classification of the new this modeling is the knowledge of linear regression, gra- objects [3]. dient methods of regression and classification, Bayesian Data prediction can be viewed as the construction of a learning, clustering, decision trees, hidden Markov mod- model to assess the class of an unlabeled sample, or to as- els, but also the theoretical knowledge about probabilistic sess the value of a target attribute that a given object is approximation learning and the Vapnik-Cervonenkis di- likely to have. In this view, classification and regression mension. Data Pattern Search, Advanced Classification are two major types of prediction problems, where classi- Methods, and Prediction Methods. fication is used to predict discrete or nominal values, while regression is used to predict continuous or ordered values. Large data processing technologies In another view, data prediction differs from the classifi- (Technical University in Košice) cation of data by the fact that the class label variable can Upon completion of the course the student will acquire acquire continuous values [1, 2, 3, 4]. basic knowledge of large data processing, methods, ap- Classification and prediction have numerous applica- proaches and technologies used in this area. From a theo- tions including credit approval, performance prediction, retical point of view the student will gain knowledge about medical diagnosis, selective marketing, prediction of wa- grid and cloud computing concepts, distributed, NoSQL ter consumption in a given territory, prediction of user rat- and in-memory database systems, parallel and distributed ing of films, recommended systems [1, 2, 3, 4]. computing methods. The student will also acquire basic skills for designing and implementing large data process- ing applications. 3 Innovation of Data Science University Subjects Knowledge discovery (Technical University in Košice) As part of the IT Academy Project, the preparing of new Basic concepts of knowledge discovery and mining in subjects and innovating of existing study materials are data. Knowledge Discovery Process, CRISP-DM method- aimed in the area of Data Science. In addition to the pre- ology, step-by-step description. Data Mining Methods - sented tasks of classification, prediction and evaluation of Predictive and Descriptive Mining in Data. Disclosure of their accuracy, many of the other topics at the university association rules, generalization, classification, prediction, subjects are part of this attractive area. A basic overview clustering. Data warehouses. Overview of selected meth- of these topics at the five partner universities of the IT ods for discovering knowledge. Mining knowledge from Academy Project is presented in the following. text document collections. Selected case studies from re- search and development projects. Logical and physical tion of information technologies and information systems, data warehouse model. Temporary and operational data specifically mediating the methodology of evaluation of repositories. Searching patterns in data. user interfaces, systems and processes. Logical and phys- ical data warehouse model. Temporary and operational Business Analyst data repositories. Data warehouses. (Technical University in Košice) Business analyst extensively uses data and advanced sta- Educational data mining tistical methods to optimize individual areas and business (Constantine the Philosopher University in Nitra) processes. The course focuses on knowledge, technology, The aim of the course is to acquaint students with com- applications and methods for continuously exploring the puter data analysis and possibilities of application of se- company’s historical performance with a view to deeper lected analytical methods in IT. Through examples and understanding and management of business planning. R- case studies, students will get acquainted with selected project as a computing environment, programming in R: data analysis methods and their applications to address data structures, graphics, applied mathematics, data anal- specific issues. Case studies are focused on the evalua- ysis. Practical examples. tion of information technologies and information systems, specifically mediating the methodology of evaluation of NoSQL databases user interfaces, systems and processes. Logical and phys- (Pavol Jozef Šafárik University in Košice) ical data warehouse model. Temporary and operational The course introduces different ways of storing and work- data repositories. Data Warehouses. Educational Data ing with large data. It compares the basic types of NoSQL Mining (EDM) can be characterized as a research area databases - DB type key-value, document DB, column DB that develops new techniques and methods, testing new and graph DB as well as practical examples of work with approaches to learning, looking for user behavior patterns selected representatives of individual types. The different of unstructured and structured data that have been created types of data representations with which these databases by interaction between different types of users in a par- work are presented. The course focuses not only on the ticular virtual learning environment, educational software, use of NoSQL databases, but also on their architecture, an intelligent or adaptive learning system, or a specialized parallel, distributed and transactional processing. software for testing. Learning Analytics (LA) is a related Programming, algorithms, complexity research area that aims to support decision-making pro- (Pavol Jozef Šafárik University in Košice) cesses at the various levels of management of an educa- The subject uses the more incoherent programming course tional organization. The objective of the course in terms that is based on the "OO-first" approach (equally starting of learning outcomes and competences. The aim of the with object-oriented programming). An important element subject is to introduce the students to the above mentioned is the visualization and use of metaphors in conjunction problems, to familiarize them with the current trends in with the graphics implemented by the JPAZ2 framework. the field of EDM and learning analytics research and to In the second part of the semester, a systematic interpre- present in practical examples the data mining methods in tation with an emphasis on a good object design takes the area of management of the educational organization. place. The course is focused on the ability to implement Searching patterns in data. Solution of practical examples. programs in the Java programming language, basic knowl- edge about the principles of object-oriented programming. Web content mining (Constantine the Philosopher University in Nitra) Basics of knowledge systems The aim of the course is to acquaint students with com- (Pavol Jozef Šafárik University in Košice) puter data analysis and possibilities of application of se- The aim of the course is to apply advanced logic courses lected analytical methods in IT. Through examples and to computer science, especially in database and knowledge case studies, students will get acquainted with selected systems. The subject of the subject is procedural seman- data analysis methods and their applications to address tics of logic programming, declarative semantics of logic specific issues. Case studies are focused on the evalua- programming and its correctness, relationship of relational tion of information technologies and information systems, database models and logic programming. specifically mediating the methodology of evaluation of user interfaces, systems and processes. Logical and phys- Data mining ical data warehouse model. Temporary and operational (Constantine the Philosopher University in Nitra) data repositories. Data Warehouses. The aim of the course The aim of the course is to acquaint students with com- is also to familiarize students with the process of discov- puter data analysis and possibilities of application of se- ering knowledge from the content of the web. Empha- lected analytical methods in IT. Through examples and sis is placed not only on text mining but also on acquir- case studies, students will get acquainted with selected ing knowledge from multimedia, searching information data analysis methods and their applications to address in document content and extracting information from text specific issues. Case studies are focused on the evalua- documents. Topics related to machine translation, natu- ral language processing and morphological analysis of the of inputs provided to customers. Work Report developer. language will be included. Participation in the project of individual actors as a func- tion of time. Output forms for the customer. Active work Neural Networks with the customer. Logical and physical data warehouse (Constantine the Philosopher University in Nitra) model. Temporary and operational data repositories. Data Neural networks represent a bio-inspired approach to warehouses. Searching patterns in data. Solution of prac- intelligent information processing computational algo- tical examples. rithms. Recognition of knowledge in memory, interpre- tation of input data, knowledge relations, generalization, Algorithms and Data Structures decision making, etc. spontaneously appear as a result of (University of Žilina) the massive parallel interaction of a large number of rela- Students will acquire basic knowledge of the theory of tively simple calculation elements. From a content point data structures and will learn how to implement them ef- of view, the subject deals with topics such as: introduction fectively. Upon completion of the course, the student is to neural networks, Binary perceptron, Backpropagation, familiar with the basic data structures and is able to use Reinforcement learning, Recurrent neural networks, their them in solving practical problems, controls the process of training and application, Self-organization. effective implementation of basic data structures. Algorithms of numerical mathematics and optimiza- Statistics for practice Advanced statistical methods tion Fundamentals of numerical methods (Constantine the Philosopher University in Nitra) (Pavol Jozef Šafárik University in Košice) The course provides an introduction to the basic methods The aim is to provide students with the theoretical knowl- of numerical mathematics and optimization. It is based on edge and practical skills needed to process real data and four aspects: mathematical basics of numerical analysis create their mathematical models using multipurpose com- and optimization, algorithmization, programming of algo- puter algebra systems, respectively. Dedicated software rithms in MATLAB and algorithm visualization. From a for addressing specific category issues. content point of view, the subject is focused on numerical methods of algebra, numerical methods of mathematical analysis and optimization methods, singleparametric op- 4 Educational tasks analysis timization methods, multiparametric optimization without The study of structures and mappings which allow to borders, least squares, and multiparametric optimization analyze the data in various forms is a challenging task. with boundaries. In this way, the first attempts to interpret the lattice Programming in Java theory as concretely as possible and to promote the better (Matej Bel University in Banská Bystrica) communication between lattice theorists and potential Program paradigms. Java programming language. JVM. users of lattice theory represent the inception for data Encapsulation. Class declaration. Instances of classes - analysis taking into account the binary relations on the objects. Constructor. Access rights and scope in class. objects and attributes sets. Since the concept hierarchies Static attributes and methods. Heredity. Exceptions. Ab- play an important role here, the term of formal concept stract classes. Polymorphism. Interface. Modularisation, analysis has been adopted for this reasoning. Briefly, abstraction, bonding, cohesion. Choice from JavaAPI. Dy- formal concept analysis scrutinizes an object-attribute namic data structures. Streams of data. Solution of practi- block of relational data in bivalent form [5]. Formal cal tasks. concept analysis allows us to explore the meaningful groupings of educational tasks (referred to objects) with Business Intelligence respect to common objectives (referred to attributes) and (Matej Bel University in Banská Bystrica) it provides the visualization capabilities. The conceptual Business Intelligence (BI) and BI type projects - focus, difficulties in mathematics education [6], or the inte- reason, goal. Repetition of necessary knowledge and skills grated care pathways [7] are analyzed by formal concept from database systems, software systems and modeling of analysis, as well. An extensive overview of the various data structures. OLTP vs. OLAP. Data warehouses, data application domains that include software mining, web warehouse development, OLAP analysis, data mining (ar- analytics, medicine, biology and chemistry data is given chitecture, metadata, implementation, ETL, OLAP, MDX, by [8, 9]. The statements that people use to communicate reports). Waterfall model working on BI project. Agile facts about the world are usually not bivalent. The truth methods of working on BI project. Processing of input in- of such statements is a matter of degree, rather than being formation from the customer. Forms of offers (customer only true or false. Fuzzy logic and fuzzy set theory are response) for the customer. Project manager role in the frameworks which extend formal concept analysis in BI project. The activity of the project architect, his com- various independent ways. In [10], our aim was to provide petence and the relevant documentation. Data Architect the system of objectives and tasks that is expected to fill and its activity in the project. ETL developer work. Types in the gap of the National Education Program in Slovak Republic. In general, the National Education Program is 11e) to apply an array data structure in a simple game pro- formulated concisely and we put emphasis on a long term gramming, to particularize other supplementary curricular documents and express the educational objectives more explicit in 12) to recognize the issues in which array data structure various areas. Therefore, we have focused on an algo- can be applied effectively, to appoint the advantages rithmic thinking area and chosen an array data structure and disadvantages of an array in comparison with as an educational content in which we have fruitfully other simple data structures (an access to elements, applied formal concept analysis. Simultaneously in this a space complexity). area, we focus on algorithms including searching, sorting The specified aims are enumerated by the revised tax- or text processing. Particularly, we have investigated onomy of Bloom in order to classify statements of what the educational tasks and objectives of five real teachers we expect or intend students to learn as a result of edu- giving lessons in computer science. We aim at specifying cation. The revised taxonomy focuses on four knowledge the particular and relatively precise objectives of an array dimensions including factual knowledge (basic elements), data structure education in the algorithmic thinking area. conceptual knowledge (interrelationships among the basic Regarding our long-term cooperation with the teachers in elements), procedural knowledge (how to do something) the field, we declare some input set of objectives of an and metacognitive knowledge (awareness and knowledge array data structure: of one’s own cognition). In general, an educational pro- cess consists of a motivation phase, a phase of the first acquisition, a fixation phase and a diagnostic phase. The 1) to specify an array as the structured homogeneous phase of a systematization, a propedeutics or an applica- data type with elements denoted by a single identi- tion phase can be also involved. fier, We submitted the previous list of aims to the teachers 2) to appoint the real examples of one-dimensional array in the secondary schools in Slovakia. The teachers were data structure (e.g. rooms in a hotel, seats in a plane, instructed to appoint the tasks which they usually apply in etc.), an educational process of an array data structure in pro- gramming. Teachers were not limited by the number of 3) to interpret the notions of an array index (an array tasks and moreover, it was possible to add some additional key) and an array element and to explain the differ- aims. Regarding five teachers data and one additional set ence between them, of 10 tasks proposed by two of the authors, we have ana- lyzed 102 tasks and 23 educational aims obtained in this 4) to distinguish an array index type and an array ele- research. We have generated the summary concept lattice ment type, and found the following observations: 5) to reason that an array index type is an ordinal type • 45 tasks (the first row in the summary concept lattice) (numbers, characters, other enumerations), are the representatives; i.e. every task includes the unique set of aims and there is no task that introduces 6) to declare a variable of array, the superset of these aims, 7) to read and to write out the array elements, • 5 tasks (from 45 representatives) are such that ev- ery task includes the unique set of aims and there is 8) to manipulate the array elements, to assign the array no task that introduces neither superset nor subset of element to the other variables, to increment the array these aims, elements, • 3 aims (the first row in the summary concept lattice) 9) to appoint the common errors related to an array data are unique, i.e. the aim is introduced only by one structure (incorrect index type, overflow, incompati- task. In effort to prepare the graduated sets of tasks, bility of the types), we have explored the longest paths extracted from the summary concept lattice with reduced labeling of all 10) to apply an array data structure in the simple issues 102 tasks. The longest path is shown in Fig. 2. Every (e.g. to store an array, to find the maximal value, to path contains the graduated system of tasks depend- modify the elements of array, etc.), ing on the final task we want to achieve in conclusion. 11a) to apply an array data structure in searching, The object label, for instance 3.5, corresponds to the fifth task of third teacher. The set of tasks labeled 6.1 11b) to apply an array data structure in sorting, - 6.10 comes from the authors. 11c) to multiply access to the array elements, The longest path illustrates that if a student has a prob- lem with Task6.5, we ask him/her to solve Task3.12. 11d) to apply an array data structure in a text processing, Moreover, if we have found that a student has a problem • Consider the starting sequence of children’s names Figure 1: The longest paths extracted from the summary and the final shift of Ferris wheel as the input. Write a concept lattice program to make a list of the children names in the se- quence in which they will get out of the Ferris wheel. • Write a program to generate twelve random values expressing the number of your website visits in a par- ticular month. Draw a histogram, highlight the max- imum and minimum and show an average value as a horizontal line. We have fruitfully applied formal concept analysis as a powerful tool in a simultaneous analysis that involves the teaching of programming skills in an open-source web ap- plication App Inventor 2. The tutorial website1 provides materials in the form of learning cards for building the ba- sic applications, but one of the authors of this paper has prepared the set of ten complex educational tasks which in summary cover 129 elements (components and their el- ements, event handlers, call, set instructions, get instruc- tions, data structures, etc.) available at the present time. The added value includes the proposal of the introductory set of complex tasks and its further modification in effort to teach and learn the different target groups. The talented lower secondary school’s pupils participate in our optional with Aim6 in Task6.5, we give him/her to fixation one task university courses and the teachers of secondary schools (or more) from the node which contains the set of equal attend the didactic workshops at our university. Our re- tasks Task1.3, Task1.4, Task1.5, Task1.8. In contrary, if sults are concerned with the inclusion of the programming a student has no problem with Task6.5, we suppose that language elements (available at the App Inventor website he/she will pass also the Task3.12. There is only one path at present) in the complex educational tasks and the effort including seven task nodes, however paths with six nodes to extract the appropriate tasks for the different types of an appear in the summary concept lattice several times. educational process. The formal context contains 10 tasks as the set objects and 129 App Inventor programming ele- Moreover, we present some interesting educational ments as the set of attributes. tasks which appear in the summary concept lattice mostly Exploring own attributes, the resulting concept lattice in the first row and one can advise them to apply in the and its attribute labels give information about the elements educational process related to an array data structure. The introduced uniquely by a particular task. As conclusion, formulations are shortened in comparison with the original we recommend the following methodology: texts. • a task with a high ratio of the own elements and the • Propose the way how to denote the parking places in low total number of elements is advised to use in a front of a hotel. How are the train carriages enumer- first acquisition phase of education, ated? How would you denote the overall and final results of six teams in the television knowledge con- • in a fixation phase, we recommend a task with a low test? ratio of the own elements and the low total number of elements, • We have observed GPS data containing ten altitudes • a task with a low ratio of the own elements and the on our tourist route. Write a program to print out the high total number of elements is suggested in a sys- altitudes on a reverse route. tematization or diagnostic educational phase, • Imagine that you have received SMS from your • a task with a high ratio of the own elements and the friend. Write a program to count the number of words high total number of elements is the least appropriate in your text message. for an educational process, because it brings many new elements without their introduction in a more • A musical instrument, like a piano, can be simulated simple task. by a computer program. Some of the keys will have assigned a particular tone frequency. Write a program to play a tone when the particular key is pressed. 1 http://www.appinventor.org/ 5 Conclusion Data can be viewed simply as the observations, entities or values which are collected in effort to form informa- tion. To ensure the readability, data used to be converted into tables and graphs. Table data appear frequently and they can be observed either directly from the medical test results, the customer habits, the scores of students or by the various transformations from the other forms, for in- stance, from database tables or questionnaires. The in- formation hidden within data can help to solve the many pending issues within community, enterprise or science. Turning of data into knowledge and wisdom is beneficial and necessary, considering either the simple computing in the spreadsheet calculators or various methods of data analysis which are more complex. Data collecting, pre- processing, reduction, visualization, dependencies explo- ration and providing the metadata are important parts of the scientific research, as well. Our ability to generate and collect data has increased rapidly in recent decades. This has prompted the need for new techniques and automated tools to help transform a large amount of data into useful information and knowl- edge. In other words, for the present time it is characteris- tic that we are rich in data, but poor in information. References [1] V. Dhar, Data science and prediction. Communications of the ACM. vol. 56, n. 12, 2013, 64-73. [2] E. Alpaydin, Introduction to Machine Learning. 2nd ed. Boston: MIT Press; 2010. [3] J. Han and M. Kamber, M., Data Mining: Concepts and Techniques, 3rd ed. Morgan Kaufmann Publishers, 2012. [4] C.O. Neil and R. Schutt, Doing Data Science: Straight Talk from the frontline. O’Reilly Media, 2014. [5] Ganter, B., Wille, R.: Formal Concept Analysis: Mathemat- ical Foundation. Springer, Heidelberg; 1999. [6] U. Priss, P. Riegler, N. Jensen, Using FCA for Modelling Conceptual Difficulties in Learning Processes, in: F. Dome- nach, D. I. Ignatov, J. Poelmans (Eds.), Contrib. to the 10th Intern. Conf. ICFCA 2012, Leuven, Belgium, 2012, 161- 173. [7] J. Poelmans, G. Dedene, G. Verheyden, H. Van der Mussele, S. Viaene, E. Peters, Combining business process and data discovery techniques for analyzing and improving integrated care pathways, in: P. Perner (Ed.): ICDM 2010, Springer- Verlag Berlin Heidelberg, 2010, 505-517. [8] J. Poelmans, D. I. Ignatov, S. O. Kuznetsov, G. Dedene, For- mal concept analysis in knowledge processing: A survey on applications, Expert Syst. Appl. vol. 40, n. 16, 2013, 6538- 6560. [9] C. Carpineto, G. Romano, Concept Data Analysis Theory and Applications, J. Wiley, 2004. [10] L’. Antoni, J. Guniš, S. Krajči, O. Krídlo, L’. Šnajder, The educational tasks and objectives system within a formal con- text. CLA 2014: 35-46.