Miloš Kudělka

Social and swarm aspects of co-authorship network

The analysis of social networks is concentrated mainly on uncovering hidden relations and properties of network nodes (vertices). Most of the current approaches are focused on different network types and different network coefficients. This article introduces a social network analysis based on the so-called Forgetting Curve and Swarm Intelligence inspired by the Ant Colony Optimization. We analyse a co-authorship network and identify two types of ties among its nodes. The Forgetting Curve and Swarm Intelligence are used to model the dynamics of such a network.

Breast Cancer Detection and Classification Using Support Vector Machines and Pulse Coupled Neural Network

This article introduces a hybrid scheme that combines the advantages of pulse coupled neural networks (PCNNs) and support vector machine, in conjunction with type-II fuzzy sets and wavelet to enhance the contrast of the original images and feature extraction. An application of MRI breast cancer imaging has been chosen and hybridization scheme have been applied to see their ability and accuracy to classify the breast cancer images into two outcomes: cancer or non-cancer. In order to enhance the contrast of the input image, identify the region of interest and detect the boundary of the breast pattern, a type-II fuzzy-based enhancement and PCNN-based segmentation were applied. Finally, wavelet-based features are extracted and normalized and a support vector machine classifier were employed to evaluate the ability of the lesion descriptors for discrimination of different regions of interest to determine whether they represent cancer or not. To evaluate the performance of presented approach, we present tests on different breast MRI images.

Local Community Detection and Visualization: Experiment Based on Student Data

This paper is focused on the detection of communities in social networks. We propose and describe a novel method for detecting local communities. We have used this method in an experiment on student social networks in order to prove our hypothesis about the nature of student communities. The results of the experiment rationalized our hypothesis and confirmed the effectiveness of the described method of local community detection.

Semantic Analysis of Web Pages Using Cluster Analysis and Nonnegative Matrix Factorization

In this paper, the web pages concerning products sale are analyzed with the aim to create clusters of similar web pages and characterize these by GUI patterns. We applied GD-CLS (gradient descent - constrained least squares) method which combines some of the best features of other methods. Both traditional methods for searching clusters and nonnegative matrix factorization are used.

Two New Methods for Network Analysis: Ant Colony Optimization and Reduction by Forgetting

This paper presents two new methods for network analysis. Ant colony optimization is a nature inspired algorithm succesfull in graph traversal and network path finding whereas network reduction based on stability introduces two new properties of network vertices based on their long-term behavior, their role in the network and the understanding of how memory works. We illustrate the algorithms on applications in social network analysis and information retrieval using the DBLP dataset and a small network of hyperlinked documents.

Visual Similarity of Web Pages

In this paper we introduce an experiment with two methods for evaluating similarity of Web pages. The results of these methods can be used in different ways for the reordering and clustering a Web page set. Both of these methods belong to the field of Web content mining. The first method is purely focused on the visual similarity of Web pages. This method segments Web pages and compares their layouts based on image processing and graph matching. The second method is based on detecting of objects that result from the user point of view on the Web page. The similarity of Web pages is measured as an object match on the analyzed Web pages.

Visualizing Clusters in the Photovoltaic Power Station Data by Sammon’s Projection

This paper presents results of the finding and the visualization cluster in the hourly recorded data of power from the small photovoltaic power station. Our main aim was to evaluate the use of Sammon’s projection for visualizing clusters in the data of power. The photovoltaic power station is sensitive for changes according to the sun’s light power. Although one can think that sunny days are the same the power of the sun light is very volatile during a day. When we wanted to analyse the efficiency of the power station, it was necessary to use some kind of clustering method. We propose the clustering method based on social network algorithms and the result is visualized by the Sammon’s projection for explorational analysis.

On spectral partitioning of co-authorship networks

Spectral partitioning is a well known method in the area of graph and matrix analysis. Several approaches based on spectral partitioning and spectral clustering were used to detect structures in real world networks and databases. In this paper, we explore two community detection approaches based on the spectral partitioning to analyze a co-authorship network. The partitioning exploits the concepts of algebraic connectivity and characteristic valuation to form components useful for the analysis of relations and communities in real world social networks.

Short Term Power Prediction of the Photovoltaic Power Station Based on Power Profiles

The use of solar energy has undergone rapid development in recent years. Photovoltaic power stations (PVPS) are often used as a source of power for smart off–grid houses. Integration of this kind of energy source is challenging because it is a source of variably generated power due to meteorological uncertainty. In this paper, we present results of the short term prediction method of generated power for small PVPS based on self–organizing maps and previously introduced power profiles.

Comparison of Local and Global Ranking in Networks

Many real world data and processes have a network structure and can usefully be represented as graphs. Network analysis focuses on the relations among the nodes exploring the properties of each network. Latest trend in analyzing networks is to focus on local methods and parallelization. We introduce a method to find the ranking of the nodes. The approach extracts dependency relations among the network’s nodes. Key technical parameter of the approach is locality. Since only the surrounding of examined nodes is used in computations, there is no need to analyze the entire network. We compare this proposed local ranking to the global ranking of PageRank. We present experiment using large-scale artificial and real world networks. The results of experiment show high effectiveness due to the locality of our approach and also high quality of node ranking comparable to PageRank.