Dzenana Donko

Determining the impact of demographic features in predicting student success in Croatia

Predicting the success of students is a topic which has been studied for a long time in different scientific fields. Evaluation of importance of the features used in the prediction and their subsequent selection is an immensely important step in the process of classification and data mining. This paper presents a study on the importance of student demographic features in the process of predicting. The study and performed analyses used the demographic data collected from the Information System for Higher Education (ISVU). For determining the importance of demographic features in the study the following methods have been used: Information Gain (IG), Gain Ratio (GR), Sequential Backward Selection (SBS), Sequential Forward Selection (SFS). The results show the features rank, their importance weight in the prediction and comparison of the results and the use of different methods. Two classification algorithms for evaluating the impact of ranking features to the quality of prediction are used: Naive Bayes i Support Vector Machine (SVM). Final results provide guidelines for the development of a new prediction model.

An Internet of Things visual domain specific modeling language based on UML

Although there are many attempts to engineer a domain specific language for the Internet of Things, most of them forget the fact that with the evolving of the Internet of Things, the end user will probably be a common person without an engineering or software development background. The designers of the UML had the same problem: how to make a language powerful enough for the professionals, but at the same time simple enough to be understood by a non-technical end user that gives the requirements. Inspired by this idea a Visual Domain Specific Modeling Language was developed for the IoT and proved that it is powerful enough for the professional and at the same time simple enough to be used by non-technical users.

An approach to design of time-aware recommender system based on changes in group user's preferences

Traditional recommender systems use collaborative filtering or content-based methods to recommend new items for users. New users and items are continuously updated to the system bringing changes in users preferences, as well as the additional context in form of temporal information. The continuous system updates change not just individual users preferences, but also group users preferences affecting prediction of ratings for individual users. In this work is presented improved user-based collaborative filtering algorithm using temporal contextual information. With difference to other approaches, we propose using weight function based on changes in the group users preferences over time that increases prediction accuracy of collaborative filtering prediction algorithm.

Predicting Quality of Web Service using IKS hybrid model

Internet and various services offered by it has become a daily routine. The Quality of Web Service (QWS) has become a significant factor in distinguishing the success of service providers. The main purpose of this paper is to analyze quality prediction using the IKS hybrid model with a new approach of data classification. We present the IKS hybrid model. The model combines selection of features, clustering and classification techniques. Three techniques are used (Information Gain (IG), K-means and Support Vector Machine (SVM)) over QWS dataset with collected 5,000 Web services. Our experiments and test results show that the proposed hybrid approach has achieved promising results in predicting the quality of web services and it represents a good basis for further development and research.

A survey of static code analysis methods for security vulnerabilities detection

Software security is becoming highly important for universal acceptance of applications for many kinds of transactions. Automated code analyzers can be utilized to detect security vulnerabilities during the development phase. This paper is aimed to provide a survey on Static code analysis and how it can be used to detect security vulnerabilities. The most recent findings and publications are summarized and presented in this paper. This paper provides an overview of the gains, flows and algorithms of static code analyzers. It can be considered a stepping stone for further research in this domain.

Temporal dynamics of changes in group user's preferences in recommender systems

Using contextual information in recommender systems is a subject of continuous improvement of rating prediction accuracy. Among others, information on temporal rating dynamics contain valuable data that establish foundation for discovering changes in both individual and group users preferences. Such changes can be caused by multiple factors such as changes of individual user interests, changes in item popularity or other hidden patterns or events. In this paper an improved user-based collaborative filtering algorithm is presented that utilizes changes of group users preferences over time. We also investigate temporal dynamics of changes in users preferences within different item categories and propose time weight function that improves prediction accuracy of recommender systems.

Feature enhanced time-aware recommender system

Traditional recommender systems utilize user and item profiles in order to predict ratings of unseen items. New users, items and ratings are continuously updated to the system, making data available for detection of changes in user preferences throughout the time. In this work the widely used user-neighborhood recommender system is extended by incorporating temporal information and enhancing measure of neighborhood similarity with information on item features. Unlike other models, we also add time-weight function in the preference prediction step to improve prediction accuracy. Experiments on real data set show an improvement in prediction performance over traditional collaborative filtering model.

Usage of data mining techniques for analyzing network intrusions

This paper presents the results of the analysis of the network intrusion detection systems using data mining techniques and anomaly detection. Anomaly detection technique is present for a while in the area of data mining. Previous papers that implement data mining techniques to detect anomaly attacks actually use well-known techniques such as classification or clustering. Anomaly detection technique combines all these techniques. They are also facing problem on the fact that many of the attacks do not have some kind of signature on network and transport layer, so it is not easy to train models for these type of attacks. Network dataset that was used in this paper is DARPA 1998 dataset created in MIT Lincoln Laboratory and is used worldwide for the network testing purposes.

Data mining meets network analysis: Traffic prediction models

Most research on network traffic prediction has been done on small datasets based on statistical methodologies. This research analyzes an internet traffic dataset spanning multiple months using the data mining process. Each data mining phase was carefully fitted to the network analysis domain and systematized in context of data mining. The second part of the paper evaluates various seasonal time series prediction models (univariate), including ANN, ARIMA, Holt Winters etc., as a data mining phase on the given dataset. The experiments have shown that in most cases ANNs are superior to other algorithms for this purpose.

Clustering approach to collaborative filtering using social networks

This paper presents results of using clustering to improve results of collaborative filtering. Clusters of users are created using friendship links within a social network using Markov Chain Algorithm (MCL). Clusters are then used to make prediction of user choices using item based collaborative filtering with cosine similarity. Using the results from analyzing different cluster sizes, new algorithm was proposed that saves time and memory resources.