Damir Pintar

The use of data mining in education environment

Modern technologies allow storage of large amounts of raw data which invariably contains usable information not yet discovered. Machine learning and data mining give us techniques which can be used to analyze that data and uncover previously unknown rules and associations, hidden knowledge which once acquired -and properly interpreted -can be used in multitude of ways. Even though data mining has mostly been used by the marketing field, lately it is not uncommon to use machine learning techniques in other environments. This paper presents how data mining algorithms and techniques can be used in the academic community to potentially improve some aspects of education quality.

Generation and Analysis of Tree Structures Based on Association Rules and Hierarchical Clustering

Detailed inspection of transactional data can reveal various useful information, in which of special importance are relationships between transaction elements. Hierarchical clustering coupled with specific distance measures reveal those relationships from one angle. Additionally, association rules - a natural method of inspecting transactional data – is able to reveal relationships between each pair of transaction elements. With transactional data modulation and multiple usages of this method a tree-like structure can be created. This paper concerns the interpretation of resulting structures for each method as well as providing comparison between them. For each algorithm mathematical base is introduced along with an explanation of result interpretation. Performances of two methods are compared and examined on a real life data set.

Integrating quantitative attributes in hierarchical clustering of transactional data

Appropriate data mining exploration methods can reveal valuable but hidden information in todays large quantities of transactional data. While association rules generation is commonly used for transactional data analysis, clustering is rather rarely used for analysis of this type of data. In this paper we provide adaptations of parameters related to association rules generation so they can be used to represent distance. Furthermore, we integrate goal-oriented quantitative attributes in distance measure formulation to increase the quality of gained results and streamline the decision making process. As a proof of concept, newly developed measures are tested and results are discussed both on a referent dataset as well as a large real-life retail dataset.

Implementation of the ebXML registry client for the ebXML registry services

The implementation of the ebXML registry services is a part of a research project at the University of Zagreb, aimed at defining a strategy for the adoption of e-business in Croatia. Implementing the ebXML registry/repository is a vital step as it implements a standard mechanism for registering, storing and retrieving data about business partners relevant for conducting e-business. This paper describes an implementation of the ebXML registry client as a diagnostic tool to test and debug the ebXML registry/repository implementation under development. The client can generate arbitrary submit objects requests to the registry services. It guarantees the conformance of the requests with the OASIS/ebXML registry services specification.

Applying the binary classification methods for discovering the best friends on an online social network

Online social networks (OSN) are one of the most widely adapted services of the Internet infrastructure, Facebook being one of the most popular among them. Facebook models connections between its users through the concept of “friendship”. However, the type and intensity of these connections between different people on Facebook vary significantly. In most cases, friends on Facebook correspond to mere acquaintances in real-life, with only a smaller subset representing actual close friends. The aim of research presented in this paper is to provide a method for estimating the intensity of Facebook friendships, i.e., to distinguish connections representing close friends from others. The study was performed by analyzing Facebook interactions between users (e.g. number of mutual likes, comments, shared photos, etc.) using supervised learning algorithms for binary classification of data. Among the chosen algorithms, the best results were gained by using random forest algorithm - accuracy of 84.73%.

The impact of training data tailoring on demand forecasting models in retail

Demand forecasting plays a very important role in retail business. Retail information systems commonly store large amounts of data which are subsequently used by sophisticated data mining tools for building forecasting models. Quality of these models is usually measured through their predictive accuracy as their most important property, followed by other measures which consider average underestimate and overestimate costs etc. Even though the choice of data mining algorithm is usually paramount, training set cleansing and preparation has a significant influence on final model performance. This article discusses and analyses the impact of training set preparation and tailoring on a final forecasting model performance used in a real world example from the retail industry.

Building an Application - Generation of �Items Tree’ Based on Transactional Data

New approach to well known data mining method of association rules is developed and elaborated. On its basic, the application that generates tree-like structure of items appearing in transactional data is created. Survey of existing applications’ functionalities using association rules are presented, along with motivation to make rather different solution. Way of interpretation of developed solution is clearly elaborated along with possible usages of developed tool and data model.

Business service interface structure

EbXML is the emerging standard intended to facilitate the efficient and flexible B2B collaboration. One of the crucial issues is the implementation of the business service interface (BSI). BSI is an intermediary between the other companies and the internal business applications of a company that is engaged in electronic business. BSI is not directly specified. Its behavior and its assignments are specified indirectly through the other specifications. This paper discusses and describes a general structure of BSI so it would comply with all the requirements which are also discussed. BSI would enable the companies to engage in electronic business collaborations according to the previously specified business processes that are agreed with the other party. It would enable conducting the business with a minimum human interference. BSI would also enable the human control of delicate business transactions.

Exploratory analysis of pairwise interactions in online social networks

ABSTRACT In the last few decades sociologists were trying to explain human behaviour by analysing social networks, which requires access to data about interpersonal relationships. This represented a big obstacle in this research field until the emergence of online social networks (OSNs), which vastly facilitated the process of collecting such data. Nowadays, by crawling public profiles on OSNs, it is possible to build a social graph where “friends” on OSN become represented as connected nodes. OSN connection does not necessarily indicate a close real-life relationship, but using OSN interaction records may reveal real-life relationship intensities, a topic which inspired a number of recent researches. Still, published research currently lacks an extensive exploratory analysis of OSN interaction records, i.e. a comprehensive overview of users’ interaction via different ways of OSN interaction. In this paper, we provide such an overview by leveraging results of conducted extensive social experiment which managed to collect records for over 3200 Facebook users interacting with over 1,400,000 of their friends. Our exploratory analysis focuses on extracting population distributions and correlation parameters for 13 interaction parameters, providing valuable insight into OSN interaction for future researches aimed at this field of study.

Automated extraction and visualization of learning concept dependencies using Q-matrices and exam results

Information systems of educational organizations often represent a potential well of useful information which can be discovered and interpreted by using specific methods. Exam results in particular are commonly used as a single-use measure of individual knowledge states, after which they are archived and subsequently never used again. Our approach suggests using past exam results as a rich data source for extracting knowledge about learning concepts, especially regarding their mutual relationships. To achieve this goal, we adopt our method for interactive visualization of patterns in transactional data and apply it to knowledge state matrices generated from real-life exam results and Q-matrices constructed by domain experts, providing the end user with rich, easily interpretable and visually engaging dendrogram structures.