Papers by Viera Rozinajova
Advances in Swarm Intelligence, 2020
Prediction of photovoltaic (PV) energy is an important task. It allows grid operators to plan pro... more Prediction of photovoltaic (PV) energy is an important task. It allows grid operators to plan production of energy in order to secure stability of electrical grid. In this work we focus on improving prediction of PV energy using nature-inspired algorithms for optimization of Support Vector Regression (SVR) models. We propose method, which uses different models optimized for various types of weather in order to achieve higher overall accuracy compared to single optimized model. Each sample is classified by Multi-Layer Perceptron (MLP) into some weather class and then model is trained for each weather class. Our method achieved slightly better results compared to single optimized model.
Springer eBooks, 2012
In this paper we deal with the interoperability of digital libraries concerned with identificatio... more In this paper we deal with the interoperability of digital libraries concerned with identification of hidden or invisible relationships within various data sources. By means of semantic processing and reasoning techniques we attempt to find the answers to sophisticated questions which are sometimes difficult also for human experts. Our initial interest was to analyze data from art museums, where we have found interesting information concerning artists, their life and work. We proceeded further to the national library databases, where we have been looking for additional information about these artists and we performed further investigation. The next step was to enrich existing records by additional useful information using web services of other libraries. By analyzing these enriched data, we could identify semantic relationships among the records, which can help us understand how these artists were influenced by each other, we can find an artist that performed in the same area as the given one, etc. This paper describes our method of processing core data, identification of semantic relationships and the experiments we have performed.
Detection of Abnormal Load Consumption in the Power Grid Using Clustering and Statistical Analysis
Lecture Notes in Computer Science, 2019
Nowadays, the electricity load profiles of customers (consumers and prosumers) are changing as ne... more Nowadays, the electricity load profiles of customers (consumers and prosumers) are changing as new technologies are being developed, and therefore it is necessary to correctly identify new trends, changes and anomalies in data. Anomalies in load consumption can be caused by abnormal behavior of customers or a failure of smart meters in the grid. Accurate identification of such anomalies is crucial for maintaining stability in the grid and reduce electricity loss of distribution companies. Smart meters produce huge amounts of load consumption measurements every day and analyzing all the measurements is computationally expensive and very inefficient. Therefore, the aim of this work is to propose an anomaly detection method, that addresses this issue. Our proposed method firstly narrows down potential anomalous customers in large datasets by clustering discretized time series, and then analyses selected profiles using statistical method S-H-ESD to calculate final anomaly score. We evaluated and compared our method to four state-of-the-art anomaly detection methods on created synthetic dataset of load consumption time series containing collective anomalies. Our method outperformed other evaluated methods in terms of accuracy.
Advances in Databases and Information Systems, 2011
Journal of Intelligent Information Systems, Mar 16, 2019
This paper presents a comparison of the impact of various unsupervised ensemble learning methods ... more This paper presents a comparison of the impact of various unsupervised ensemble learning methods on electricity load forecasting. The electricity load from consumers is simply aggregated or optimally clustered to more predictable groups by cluster analysis. The clustering approach consists of efficient preprocessing of data obtained from smart meters by a model-based representation and the K-means method. We have implemented two types of unsupervised ensemble learning methods to investigate the performance of forecasting on clustered or simply aggregated load: bootstrap aggregating based and the newly proposed density-clustering based. Three new bootstrapping methods for time series analysis methods were newly proposed in order to handle the noisy behaviour of time series. The smart meter datasets used in our experiments come from Australia, London, and Ireland, where data from residential consumers were available. The achieved results suggest that for extremely fluctuating and noisy time series the forecasting accuracy improvement through the bagging can be a challenging task. However, our experimental evaluation shows that in most of the cases the density-based unsupervised ensemble learning methods are significantly improving forecasting accuracy of aggregated or clustered electricity load.
Improving Time Series Prediction via Modification of Dynamic Weighted Majority in Ensemble Learning
Lecture Notes in Computer Science, 2018
In this paper, we explore how the modified Dynamic Weighted Majority (DWM) method of ensemble lea... more In this paper, we explore how the modified Dynamic Weighted Majority (DWM) method of ensemble learning can enhance time series prediction. DWM approach was originally introduced as a method to combine predictions of multiple classifiers. In our approach, we propose its modification to solve the regression problems which are based on using differing features to further improve the accuracy of the ensemble. The proposed method is then tested in the domain of energy consumption forecasting.
Prediction of Power Load Demand Using Modified Dynamic Weighted Majority Method
Advances in intelligent systems and computing, Nov 5, 2016
The paper deals with the prediction of electricity demand, using data from smart meters obtained ... more The paper deals with the prediction of electricity demand, using data from smart meters obtained in defined time steps. We propose the modification of ensemble learning method called Dynamic Weighted Majority (DWM). The data are represented by data streams. According to our experiments, the proposed solution offers favorable alternative to current solutions. We also focus on the comparison of proposed ensemble method with single predictions used in the model.
Intelligent Support for Program Development
PPIG, 1996
EduVirtual - Modern Educational Platform based on Multimedia Technologies
2019 International Symposium ELMAR
This paper presents an innovative project in the field of education, which applies modern graphic... more This paper presents an innovative project in the field of education, which applies modern graphic technologies, such as virtual and augmented reality. The project consists of multiple applications, which are using various technologies for different learning purposes. The applications are connected through user profiles and results are sent to a central server where a teacher can see all students’ recordings. The goal of the project is to deliver a solution that is easy to use without technical knowledge and is also engaging for students in primary and high schools. This solution should also provide an easy way for teachers to create new educational content and to monitor and analyse students’ progress.
Prediction of electricity consumption using biologically inspired algorithms
2017 IEEE 14th International Scientific Conference on Informatics, 2017
Prediction of electricity consumption has become a very investigated field of research recently. ... more Prediction of electricity consumption has become a very investigated field of research recently. By lowering prediction error, we can minimize costs of suppliers. The classical approach using one prediction model has been proved insufficient. The problem is that none of the currently existing prediction models is sufficiently robust to estimate the forecast of time series accurately. One of the solutions to this problem is Ensemble learning. This approach attempts to create the best possible prediction by employing several different forecasting methods. We focused particularly on the ways of combining these methods. We apply different swarm intelligence algorithms to find the optimal combinations. The aim of this study is to compare prediction accuracy of the ensemble learning model to base forecasting methods. We also examine the optimal size of forecasting methods in the ensemble.
Building an Agent for Factual Question Generation Task
2018 World Symposium on Digital Intelligence for Systems and Machines (DISA), 2018
With the boom of e-learning and online education systems, also question generation systems have b... more With the boom of e-learning and online education systems, also question generation systems have become more interesting. Nowadays, it is relatively common and simple to use internet and web technologies for learning. There are many sources of educational materials in various formats like text, audio, video or interactive applications and games. But the automatic evaluation and confirmation of students' knowledge is still difficult. In this paper, we describe our endeavour to design and create an interactive educational agent which to some extent acts as a teacher: it automatically generates factual questions from the educational text and tries to reveal if the student understood the information presented there. We have evaluated our agent in comparison with state-of-the-art question generation systems. In order to assess the applicability of generated questions in educational process we have also compared these questions with those expected by the students.
Transactions on Large-Scale Data- and Knowledge-Centered Systems L, 2021
]. This version is published under a Creative Commons CC-BY-NC-ND. No commercial redistribution o... more ]. This version is published under a Creative Commons CC-BY-NC-ND. No commercial redistribution or re-use allowed. Derivative works cannot be distributed. © Matej Kloska, Viera Rozinajova.
Incremental Time Series Prediction Using Error-Driven Informed Adaptation
2016 IEEE 16th International Conference on Data Mining Workshops (ICDMW), 2016
This paper presents an approach to predictive modeling of sequentially arriving data, also known ... more This paper presents an approach to predictive modeling of sequentially arriving data, also known as a stream. Because of their unique properties, this kind of data requires different mining techniques. The ultimate limitations are the memory and the time. Since the number of records can be infinite, it is not possible to store them all in memory or read them more than once. Hence, the prediction method should work incrementally. Another important aspect of these data is that their characteristics change over time. The identification of the ongoing change in the monitored data sequence, also called "concept drift", can significantly help to improve prediction accuracy by prediction model adaptation to the drifts. The challenge is to perform these model adaptations online. We have proposed an incremental adaptive method for time series prediction. Our approach is based on the adaptive learning scheme "predict – diagnose – update". The main concern was to find out whether our error-driven informed adaptation approach can equal the traditional blind adaptation approaches in accuracy and required resources such as time and memory. The results showed that informed adaptation can achieve comparable accuracy but uses less computational resources. We focused specifically on power demand forecasting but we showed that the approach is applicable also on time series with similar characteristics from other domains.
2012 Federated Conference on Computer Science and Information Systems, 2012
Service oriented architecture (SOA) is nowadays one of the dominant styles in developing new info... more Service oriented architecture (SOA) is nowadays one of the dominant styles in developing new information systems. These information systems often have complex models, which can contain mistakes, or are described by informally. In order to minimize mistakes and to create formal models, patterns as components of software development could be used-according to Model driven development (MDD) principles. Design patterns in SOA have been identified by T. Erl [1]. However, they are represented in form which is suitable for humans, but not for computers. In context of machine processing formal representation of patterns would be advantageous. In this paper we present our approach to partial formal representation of SOA design patterns using production rules. This partial formal representation is useful in searching for mistakes (antipatterns) in models and will enable creating formal models (with patterns) from informal documents.
SOA Design Patterns -- Can They Improve the Process  of Model Driven Development?
2013 IEEE International Conference on Services Computing, 2013
SOA brought significant progress in information systems development, but the full exploitation of... more SOA brought significant progress in information systems development, but the full exploitation of its advantages is sometimes hindered by specific problems. One of them is concerned with paradigm of Model Driven Development (MDD), where it is necessary to control correctness of the whole design within many models on different levels of abstraction. This is the place, where we would like to contribute. We propose a new approach to application of SOA design patterns in the development process. Patterns are transformed into a machine acceptable form using category theory and attributed graphs. This form is language independent, leads up to automated detection of pattern or antipattern in the description/model of SOA based system and in the future can facilitate also correctness checking.
Software Service Recommendation Using Context
2015 4th Eastern European Regional Conference on the Engineering of Computer Based Systems, 2015
Nowadays a great number of different software services are available for usage. These services ha... more Nowadays a great number of different software services are available for usage. These services have different attributes and proper selection of software service for solving given problem is very difficult and non-trivial task. This paper focuses on recommendation of software services. We want to achieve selection of software services which fulfills needs of a user to the greatest possible extent. We focus on hybrid approach to recommendation, while using contextual information in the process. Context is used for representing information about each user and service separately. We also propose a method for context refinement with the expectation of retrieving more accurate results in future recommendations.
Journal of King Saud University - Computer and Information Sciences, 1996
The paper investigates possibilities of writing programs with having the relevant knowledge on pr... more The paper investigates possibilities of writing programs with having the relevant knowledge on programing available in explicity fonn. in order to perform experiments, a knowledge base was built which codes some of the knowledge related to the problem of selecting a proper data type in the process of program formation. The base is presented in paper along with one experiment which also shows the system performance and user-system interaction. The experiment setting is such that the llser makes a guess which data type is appropriate to use and this hypothesis is either confirmed or rejected by the system. Moreover, as a result of the system's deductive reasoning other acceptable data types are proposed by the system. The result shows that the system is able to provide an advice to a programer. This can be particularly useful in the process of learning programing.
Identifying Semantic Relationships in Digital Libraries of Cultural Heritage
Lecture Notes in Computer Science, 2012
In this paper we deal with the interoperability of digital libraries concerned with identificatio... more In this paper we deal with the interoperability of digital libraries concerned with identification of hidden or invisible relationships within various data sources. By means of semantic processing and reasoning techniques we attempt to find the answers to sophisticated questions which are sometimes difficult also for human experts. Our initial interest was to analyze data from art museums, where we have found interesting information concerning artists, their life and work. We proceeded further to the national library databases, where we have been looking for additional information about these artists and we performed further investigation. The next step was to enrich existing records by additional useful information using web services of other libraries. By analyzing these enriched data, we could identify semantic relationships among the records, which can help us understand how these artists were influenced by each other, we can find an artist that performed in the same area as the given one, etc. This paper describes our method of processing core data, identification of semantic relationships and the experiments we have performed.
informatik.rwth-aachen.de
Abstract: In this paper we deal with design and creation of web mashups which represent one of th... more Abstract: In this paper we deal with design and creation of web mashups which represent one of the important Web 2.0 application approaches. The main goal of this work is to describe an approach where a recommendation about existing and cooperating data ...
Intelligent Support to Program Development
Uploads
Papers by Viera Rozinajova