Michał Kozielski

DISESOR - decision support system for mining industry

This paper presents the DISESOR integrated decision support system. The system integrates data from different monitoring and dispatching systems and contains such modules as data preparation and cleaning, analytical, prediction and expert system. Architecture of the system is presented in the paper and a special focus is put on the presentation of two issues: data integration and cleaning, and creation of prediction model. The work contains also a case study presenting an example of the system application.

Regression Rule Learning for Methane Forecasting in Coal Mines

The rule-based approach to methane concentration prediction is presented in this paper. The applied solution is based on the modification called fixed of the separate-and-conquer rule induction approach. We also proposed the modification of a rule quality evaluation based on confidence intervals calculated for positive and negative examples covered by the rule. The characteristic feature of the considered methane forecasting model is that it omits the readings of the sensor being the subject of forecasting. The approach is evaluated on a real life data set acquired during a week in a coal mine. The results show the advantages of the introduced method (in terms of both the prediction accuracy and knowledge extraction) in comparison to the standard approaches typically implemented in the analytical tools.

Multilevel Conditional Fuzzy C-Means clustering of XML documents

XML documents are the special kind of data having hierarchical structure. Typical clustering algorithms do not meet requirements which may be stated for analysis of such data. A novel, dedicated for XML documents clustering method called Multilevel clustering of XML documents (ML) is presented in the paper. The method clusters feature vectors encoding XML documents on the different structure levels. Application of Conditional Fuzzy C-Means algorithm to ML method is proposed in the paper and the advantage of this fuzzy method over hard approach to ML algorithm is discussed and proved. An application of ML method to accelerating query execution on XML documents is discussed in the paper. The experimental results performed on two data sets having different characteristics show that the proposed method of multilevel conditional fuzzy clustering of XML documents outperforms hard multilevel clustering.

Evaluation of semantic term and gene similarity measures

In this paper we present the results of the research verifying how the functional description of genes contained in Gene Ontology database is related to genes expression values recorded during biological experiments. We compare several different gene similarity measures and semantic term similarity measures, and evaluate how the similarity of genes based on Gene Ontology terms is correlated with similarity of genes based on expression profiles. The analysis are preformed on three different datasets and we show that there is no single term similarity measure that always gives the best correlation results. The choice of the best term similarity measure depends on dataset characteristic.

Fuzzy Clustering and Gene Ontology Based Decision Rules for Identification and Description of Gene Groups

The paper presents results of the research verifying whether gene clustering that takes under consideration both gene expression values and similarity of GO terms improves a quality of rule-based description of the gene groups. The obtained results show that application of the Conditional Robust Fuzzy C-Medoids algorithm enables to obtain gene groups similar to the groups determined by domain experts. However, the differences observed in clustering influences a description quality of the groups. The rules determined cover more genes retaining their statistical significance. The rules induction and post-processing method presented in the paper takes under consideration, among others, a hierarchy of GO terms and a compound measure that evaluates the generated rules. The approach presented is unique, it makes possible to limit a number of rules determined considerably and to obtain rules that reflect varied biological knowledge even if they cover the same genes.

Improving the Results and Performance of Clustering Bit-encoded XML Documents

Clustering XML documents according to their structure is one of the techniques that may improve the effectiveness of XML documents storage and retrieval. One of existing approaches to this problem is to encode XML document structure as a string of bits and cluster such feature vectors. High dimensionality and sparseness of the feature vectors are the weaknesses of this method. The paper presents four methods reducing the dimensionality of the bit feature vectors. Two of these methods are novel. They are dedicated to XML documents and should be applied during the encoding process. The results showed good efficiency of these inner-encoding methods and their ability of improving clustering results in some cases. The methods presented in the paper are tested on two datasets of XML documents having different characteristics

Soft Approach to Identification of Cohesive Clusters in Two Gene Representations

Abstract The approach to identify clusters of genes represented both by expression values and Gene Ontology annotations, where cluster membership should not be in conflict with any of the representations is presented in the paper. The method enables to identify the genes that are differently clustered in different representations, what can lead to further analysis and interesting conclusions. The approach is based on the fuzzy clustering algorithms and the notion of proximity as the aggregation operation at the higher level than similarity matrices is performed. The approach is verified on two datasets: a small synthetic and real-world gene dataset.

Correlation of Genes Similarity Measures Based on GO Terms Similarity and Gene Expression Values

In this paper we present results of analysis if (and how) the functional similarity of genes can be compared to the similarity resulting from raw experimental data. We assume that information provided by Gene Ontology database can be regarded as an expert knowledge on genes and their function and therefore it should be correlated with genes similarity obtained based on analysis of raw expression data. We analyse several different measures of genes similarities in the Gene Ontology (GO) domain and compare the obtained results with the genes similarities observed in the expression level domain. We perform the analysis on three datasets on different characteristics. We shows that there is no single measure which gives the best results in all cases, and the choice of appropriate gene similarity measure depends on sets characteristics. In most cases, the best results are obtained by Avg-sum gene similarity measure in combination with Path–length GO terms similarity measure.

The Diffusion of Viral Content in Multi-layered Social Networks

Modelling the diffusion of information is one of the key areas related to activity within social networks. In this field, there is recent research associated with the use of community detection algorithms and the analysis of how the structure of communities is affecting the spread of information. The purpose of this article is to examine the mechanisms of diffusion of viral content with particular emphasis on cross community diffusion.

Integration of Facebook Online Social Network User Profiles into a Knowledgebase

The article describes attempts made to integrate variety of user’s data available on Facebook online social network. The source of the data are user profiles, publicly available for other Facebook users, which contain data such as visited places, favorite sport teams, TV programs, watched movies, read books and other likes. The destination of integrated data is FOAF ontology adopted for integration purposes. The work presents the required FOAF ontology extensions and an approach to Facebook data extraction as a contribution. Also the query and reasoning examples on the created knowledgebase are presented.