Eleonora Brtka

A comparison of rule sets induced by techniques based on rough set theory

The main goal of this paper is to investigate the performance of rule sets obtained by different techniques based on the rough set theory. The performance of a rule set is defined as its ability to classify a set of objects into predefined classes. The set of rules contains the if then form rules. The set of objects is represented by flat table (table, organized data) where each object is represented by its attributes. To generate the rule sets we have used Rosetta software system. The performance of rule sets is measured by employment of the confusion matrix.

The data visualization technique in e-learning system

Data mining techniques and their applications are widely recognized as powerful tools in various domains. Educational data mining is an emerging discipline, concerned with developing methods for exploring the unique types of data that come from the educational context. In the domain of education, there are varieties of data of different types collected during the educational process. The main question is: Is it possible to process the collected data with the data mining system and what are main advantages of data mining and e-learning interaction? In this paper, we present an insight into the possible interaction between course management system and data mining techniques. The main goal of this work is to investigate some data mining techniques in order to deliver most appropriate learning object to the learner. In this paper, visualization as a data mining technique is investigated. For this purpose, free data mining tool Weka was used.

The representation of indiscernibility relation by graph

The paper presents a new form of indiscernibility relation based on graph. Based on widely accepted definitions of indiscernibility relation and its matrix representation, it has been shown how the indiscernibility relations can be obtained by a graph. The application of the indiscernibility graph enables the partitioning of the universe of objects represented by their attributes. This is in connection with the Rough sets theory (Z. Pawlak, 1980s). The relation between state space, when applied to the indiscernibility graph, is also theoretically discussed. Formal definitions are accompanied by some examples.

Application of data mining techniques to regulated learning system

This paper investigates some theoretical aspects of so called educational data mining, its characteristics and performances. It is well known that different Course Management Systems — CMS are widely available. They can store a vast amount of information about learners and learning objects. It is also well known that the estimation of conformity of learning objects to the learner personal profile is very important. A vast amount of data about learners and learning objects cannot be inspected manually. That is why the usage of a data mining system is necessary in order to gain information “hidden” in data bases created by CMS. The main goal of this work is to investigate some data mining techniques in order to deliver most appropriate learning object to the learner. The proposed application of educational data mining techniques is followed by a case study.

Effects of the distribution of the values of condition attribute on the quality of decision rules

The table-organized data can be analyzed by various algorithms; some of them are capable of generating IF THEN decision rules which comprises of condition attributes and decision attributes. However, it is possible to reduce the set of condition attributes but without information loss. By analysis of the condition attributes set and cuts histogram obtained by discretization and rule consistency, it is possible to choose condition attributes. This paper gives some directions and the practical example.

Improving Readability of Medical Data by Using Decision Rules

As medical journal abstracts have become more and more difficult to read, there is a burning issue for doctors to get relevant medical information in order to solve a problem in a fast and efficient way. The paper deals with the synthesis of sentences of spoken language from tabular historical data that relate to a specific medical sub domain. In this case Systematic Syntax Classification of Objects or SSCO algorithm was used in order to generate decision rules which were consequently transformed to natural language and delivered to the user by a machine text reader. The system is “hands–free”, reliable, and enables communication by natural language. The experiments were conducted on data sample consisting of patient’s conditions after hip surgery procedure and originating from General hospital “Djordje Joanovic”, Zrenjanin, Serbia.

The rough sets based data analysis in small and medium sized enterprises

This paper deals with techniques of data analyses based on the rough sets theory and similarity relations. Local and global similarity relations have been described and used to analyze data. The process of the aggregation of local similarities has also been described. The technique is very useful when data is collected by various small and medium-sized enterprises. It is well known that there is a growing need for developing a systematic approach in establishing an insight to the success rate of services offered by small and medium-sized enterprises. An experiment is conducted on small data sample available via Internet ftp protocol. Results show how these techniques can be implemented and used in various enterprises.

The cognitive conflict in the extended RED - PiE model

This paper deals with an extended RED –PiE model (Dix, 1991), named RECgD –PiE model which incorporates the cognitive aspects in the domain of education. Special interest of this work is so called cognitive conflict which is familiar in the domain of education. It is proposed to form the set of If - Then rules in order to describe model behavior including its cognitive component. The preliminary experiment is conducted and it is concluded that inclusion of the set of If Then rules is possible into the software application based on RECgD –PiE model.