Ing. Jaroslav Kuchař, Ph.D.

Článek

10.2807/1560-7917.ES.2022.27.33.2200070

Katedra softwarového inženýrství

Background Analyses of diagnostic performance of SARS-CoV-2 antigen rapid diagnostic tests (AG-RDTs) based on long-term data, population subgroups and many AG-RDT types are scarce. Aim We aimed to analyse sensitivity and specificity of AG-RDTs for subgroups based on age, incidence, sample type, reason for test, symptoms, vaccination status and the AG-RDT’s presence on approved lists. Methods We included AG-RDT results registered in Czechia’s Information System for Infectious Diseases between August and November 2021. Subpopulations were analysed based on 346,000 test results for which a confirmatory PCR test was recorded ≤ 3 days after the AG-RDT; 38 AG-RDTs with more than 100 PCR-positive and 300 PCR-negative samples were individually evaluated. Results Average sensitivity and specificity were 72.4% and 96.7%, respectively. We recorded lower sensitivity for age groups 0–12 (65.5%) and 13–18 years (65.3%). The sensitivity level rose with increasing SARS-CoV-2 incidence from 66.0% to 76.7%. Nasopharyngeal samples had the highest sensitivity and saliva the lowest. Sensitivity for preventive reasons was 63.6% vs 86.1% when testing for suspected infection. Sensitivity was 84.8% when one or more symptoms were reported compared with 57.1% for no symptoms. Vaccination was associated with a 4.2% higher sensitivity. Significantly higher sensitivity levels pertained to AG-RDTs on the World Health Organization Emergency Use List (WHO EUL), European Union Common List and the list of the United Kingdom’s Department of Health and Social Care. Conclusion AG-RDTs from approved lists should be considered, especially in situations associated with lower viral load. Results are limited to SARS-CoV-2 delta variant.

Článek

10.32614/RJ-2019-048

Katedra softwarového inženýrství

Several methods for creating classifiers based on rules discovered via association rule mining have been proposed in the literature. These classifiers are called associative classifiers and the best-known algorithm is Classification Based on Associations (CBA). Interestingly, only very few implementations are available and, until recently, no implementation was available for R. Now, three packages provide CBA. This paper introduces associative classification, the CBA algorithm, and how it can be used in R. A comparison of the three packages is provided to give the potential user an idea about the advantages of each of the implementations. We also show how the packages are related to the existing infrastructure for association rule mining already available in R.

Stať ve sborníku

Katedra softwarového inženýrství

With the availability of the vast amount of users and Location-based social networks, the problem of POI recommendations has been widely studied and received significant research attention in the last years. While previous works of POI recommendation mostly focused on investigating the spatial, temporal, and social influence, the use of additional content information has not been directionally studied. In this paper, we propose the content-aware matrix factorization method based on incorporating POI attributes and categories information. We propose two variants of the algorithm that can work with an explicit and implicit feedback. Experimental results show that the proposed method improves the quality of recommendation and outperforms most state-ofthe-art collaborative filtering algorithms.

Stať ve sborníku

Katedra softwarového inženýrství

Senzorická síť veřejného osvětlení na pražském Karlínském náměstí poskytuje měření znečištění ovzduší polétavým prachem PM10 jako otevřená data. V této práci v nich detekujeme anomálie pomocí algoritmů strojového učení pro predikci časových řad a prahování. Chceme, aby se algoritmus strojového učení naučil pravidelnosti v datech a pokud se stane něco neočekávaného, tak to prahováním odhalíme. Experimentovali jsme s lineární regresí a LSTM rekurentní neuronovou sítí, které jsme mezi sebou porovnávali střední kvadratickou chybou. Ukázalo se, že lineární regrese, která predikuje z posledních dvou měření, dosahuje lepších výsledků. Anomálie jsme detekovali z rozdílů predikovaných a skutečných hodnot. Práh pro detekování anomálií jsme vypočítali z histogramu rozdílů predikcí a skutečně naměřených hodnot. Testování ukázalo, že takto navržená metoda dokáže odhalit některé anomálie v měřeních polétavého prachu PM10, ale mnoho anomálií (například postupně nabíhajících) nedetekuje.

Stať ve sborníku

Katedra softwarového inženýrství

Classification models composed of crisp rules provide excellent explainability. The limitation of many conventional rule learning algorithms is the separate-and-conquer strategy, which may be slow on large data. Association Rule Classifiers (ARC) is an alternative approach that can be very fast on massive datasets but is highly susceptible to the correct choice of metaparameters. Most existing ARC algorithms use default thresholds of 50% for minimum confidence and 1% minimum support, which can result in excessively long rule generation or underperforming models. Due to the high-costs that can be associated with evaluation of single combination, it is impractical to use standard metaparameter optimization approaches. In this paper, we introduce two variant threshold tuning algorithms specifically designed for ARC. Evaluation on 22 standard UCI datasets shows promising results in terms of model size and accuracy in comparison with the default thresholds. The implementation of the proposed algorithms is made available in R packages rCBA and arc, which are available in the CRAN repository.

Článek

10.18267/j.aip.114

Katedra softwarového inženýrství

Metody dolování z dat jsou aplikovány ve stále větší míře, a to i v doménách, které tradičně nemají tak silnou podporu analytických nástrojů a kde převládá ruční práce analytika. Použití těchto metod v oblasti fiskálních dat umožní jejich hlubší analýzu a může přinést nová zjištění. Nasazení pokročilých metod dolování z dat je jednou z částí projektu OpenBudgets.eu, který se zaměřuje na transparentnost a odpovědnost v oblasti nakládání s veřejnými prostředky. Tento přehledový článek shrnuje některé zkušenosti autorů z tohoto projektu získané při vývoji, implementaci a aplikaci vybraných metod dolování z fiskálních dat. Jedná se zejména o metody detekce anomálií a dolování asociačních pravidel. Tyto metody jsou integrovány do centrální platformy projektu, která je k dispozici pokročilým i běžným uživatelům v případě zájmu o analýzu fiskálních dat. Pilotní analýzy ukázaly, že problémem dataminingové analýzy v této doméně je velký objem nacházených pravidel a různorodý původ jejich vzniku.

Článek

10.1016/j.knosys.2018.03.006

Katedra softwarového inženýrství

EasyMiner (http://www.easyminer.eu) is a web-based machine learning system for interpretable machine learning based on frequent itemsets. The system currently offers association rule learning (apriori, FP-Growth) and classification (CBA). For association rule learning and classification, EasyMiner offers a visual interface designed for interactivity, allowing the user to define a constraining pattern for the mining task. The CBA algorithm can also be used for pruning of the rule set, thus addressing the common problem of “too many rules” on the output, and the implementation supports automatic tuning of confidence and support thresholds. The development version additionally supports anomaly detection (FPI and its variations) and linked data mining (AMIE+). EasyMiner is dockerized, some of its components are available as open source R packages.

Stať ve sborníku

Katedra softwarového inženýrství

This work is a brief summary of a master thesis that focuses on design and implementation of a framework that uses computers of website visitors as computing nodes through web browsers. It contains an analysis of the Web environment, summarization of previous approaches and projects, design and implementation of the framework. The work describes the solution of computing node failure, reaction to slow computing node, possibilities of controlling the load of the framework on a website visitor’s computer, strategies for work distribution and security of the framework. At the end of the work, the experiment results and proposal of improvements are listed.

Stať ve sborníku

Katedra softwarového inženýrství

Approaches for the anomaly detection task based on frequent pattern mining follow the paradigm: if an instance contains more frequent patterns, it means that this data instance is unlikely to be an anomaly. This concept can be used in financial industry to reveal contextual anomalies. The main contribution of this paper is an approach that includes a novel formula for computation of anomaly scores. We evaluated the proposed approach on baseline datasets and present a use case on a real world financial dataset. We also propose a way how to explain the anomaly to the users. Implementations of the evaluated algorithms and experiments are available online in R.

Stať ve sborníku

Katedra softwarového inženýrství

Aplikace pro mobilní telefony se staly každodenní součástí našich životů. Jejich pole působnosti je ale zatím ve většině případů striktně limitované displejem daného zařízení. Jedním ze současných trendů ve vývoji mobilních aplikací je právě rozšiřování možností aplikace za hranice samotného přístroje -- do prostředí virtuální, potažmo rozšířené reality. Cílem tohoto projektu je vytvoření systému, který umožní obohacení mobilních aplikací o vyhledávání v prostředí rozšířené reality. Hlavní komponentou systému je serverová aplikace poskytující službu k vyhodnocení shod přijatého obrázku s definovanou databází obrázků. Pro usnadnění implementace vyhledávání v rozšířené realitě je k dispozici knihovna pro platformu iOS. Správa databází obrázků je umožněna prostřednictvím REST API nebo pomocí jednoduchého webového rozhraní.

Stať ve sborníku

Katedra softwarového inženýrství

EasyMiner (easyminer.eu) is an academic data mining project providing data mining of association rules, building of classification models based on association rules and outlier detection based on frequent pattern mining. It differs from other data mining systems by adapting the “web search” paradigm. It is web-based, providing both a REST API and a user interface, and puts emphasis on interactivity, simplicity of user interface and immediate response. This paper will give an overview of research related to the EasyMiner project.

Článek

10.1016/j.knosys.2017.07.026

Katedra softwarového inženýrství

Interest Beat (inbeat.eu) is an open source recommender framework that fulfills some of the demands raised by emerging applications that infer ratings from sensor input or use linked open data cloud for feature expansion. As a recommender algorithm, InBeat uses association rules, which allow to explain why a specific recommendation was made. Due to modular architecture, other algorithms can be easily plugged in. InBeat has a pure JavaScript version, which allows to confine processing to a client-side device. There is a performance optimized server-side bundle, which succesfully participated in two recent recommender competitions involving large volumes of streaming data. InBeat works on a number of platforms and is also available for Docker.

Stať ve sborníku

Katedra softwarového inženýrství

Digital editions of newspapers cause information overflow and users have problems choosing what they want to read. Systems which recommend news articles are suitable to solve such problems. Nevertheless, they face challenges unknown to the systems recommending books or movies such as a frequency of producing the new content. CLEF NewsREEL challenge enables to compare and evaluate news recommendation systems in an online and offline task focused on recommending articles to real users and tuning of algorithms respectively. This paper deals with an approach based on association rules acting as a classifier. In our approach we experimented with settings that allows to reduce the amount of rules used for the classification and increase the performance that is crucial for real recommendations. We evaluated our approach in both tasks of the CLEF NewsREEL 2017 challenge.

Stať ve sborníku

Katedra softwarového inženýrství

PMML is an industry-standard XML-based open format for representing statistical and data mining models. Since PMML does not yet support outlier (anomaly) detection, in this paper we propose a new outlier detection model to foster interoperability in this emerging field. Our proposal is included in the PMML RoadMap for PMML 4.4. We demonstrate the proposed format on one supervised and two unsupervised outlier detection approaches: association rule-based classifier CBA, frequent-pattern based method FPOF and isolation forests.

Stať ve sborníku

Katedra softwarového inženýrství

In this paper we summarize our experiments with a rule-based classifier as a recommender within CLEF NewsREEL 2017 challenge. Systems that recommend news articles are suitable to solve information overflow in digital editions of newspapers, when users have problems choosing what they want to read. They face challenges unknown to the systems recommending books or movies such as a frequency of producing the new content. This paper deals with an approach based on association rules acting as a classifier. In our approach we experimented with settings that allow reducing the amount of rules used for the classification and increasing the performance that is crucial for real recommendations.

Stať ve sborníku

Katedra softwarového inženýrství

This paper presents a use case for the data mining system EasyMiner in European project OpenBudgets.eu, which is concerned with publication and analysis of financial data of municipalities. EasyMiner is a web-based data mining system. This paper focuses on its new outlier detection functionality, which relies on frequent pattern mining. In addition, the system supports association rule discovery and building of rule-based classification models. The system exposes a REST API and can thus be easily integrated in third party applications.

Stať ve sborníku

Katedra softwarového inženýrství

V souvislosti s rostoucí popularitou využívání data miningových dat lze registrovat také rostoucí poptávku po možnosti integrace data miningových algoritmů a systémů do komplexnějších, uživatelsky přívětivějších aplikací. Tento příspěvek prezentuje novou verzi systému EasyMiner, integrovanou do softwarového řešení vyvíjeného v rámci evropského projektu OpenBudgets.eu, který je zaměřen na zpřístupňování a analýzy finančních dat samospráv. EasyMiner je webový data miningový systém podporující dolování asociačních pravidel, tvorbu klasifikačních modelů a v současné verzi nově také detekci outlierů. Příslušná funkcionalita je k dispozici nejen prostřednictvím grafického uživatelského rozhraní, ale také prostřednictvím komplexního REST API.

Stať ve sborníku

Katedra softwarového inženýrství

V dnešní době, kdy množství informací na internetu stále narůstá, se automatické zpracování a třídění dat stalo velmi oblíbeným oborem informačních technologií. Jednou z oblastí je i internetové zpravodajství. Cílem tohoto projektu je nástroj pokrývající celý proces pro základní analýzu článků z českých zpravodajských serverů. Projekt je zaměřen především na extrakci relevantních dat a jejich analýzu. V první části zahrnuje ale i související crawler, díky kterému je možné stáhnout články k analýze ze zpravodajských webů. V druhé části je ze stažených HTML stránek automaticky extrahován relevantní obsah článků a jejich další atributy. Třetí částí je pak textová analýza využívající existující postupy a nástroje, která se zaměřuje na extrakci pojmenovaných entit a analýzu sentimentu českého textu. Nad výslednými strukturovanými daty se lze dotazovat z různých pohledů a provádět tedy různé druhy experimentů.

Stať ve sborníku vyzvaná či oceněná

10.1007/978-3-319-30996-5_11

Katedra softwarového inženýrství

In the recent years, there is a significant interest in a link prediction - an important task for graph-based data structures. Although there exist many approaches based on the graph theory and factorizations, there is still lack of methods that can work with multiple types of links and temporal information. The creation time of a link is an important aspect: it reflects age and credibility of the information. In this paper, we introduce a method that predicts missing links in RDF datasets. We model multiple relations of RDF as a tensor that incorporates the creation time of links as a key component too. We evaluate the proposed approach on real world datasets: an RDF representation of the ProgrammableWeb directory and a subset of the DBpedia focused on movies. The results show that the proposed method outperforms other link prediction approaches.

Stať ve sborníku

Katedra softwarového inženýrství

Webová aplikace EasyMiner je akademický nástroj pro získávání znalostí z malých a středně velkých dat ve formě asociačních pravidel. Nová verze tohoto systému využívá prostředí Apache Hadoop a Apache Spark pro zpracování velkých datových zdrojů na výpočetním clusteru MetaCentra sdružení CESNET. Aplikace se skládá z několika mikro služeb, které se starají o nahrávání velkých dat do distribuovaného úložiště HDFS, transformaci dat v clusteru do normalizované formy a dolování znalostí z datasetů v podobě asociačních pravidel s využitím výpočetních prostředků clusteru pomocí nástroje Apache Spark. S těmito mikro službami se dá komunikovat prostřednictvím RESTového rozhraní a jako celek tvoří data miningový software fungující jako webová služba - SaaS.

Stať ve sborníku

Katedra softwarového inženýrství

Augmenting a feature set using mappings to the Web of data is an up-and-coming way to enrich data in the original dataset. Those enrichments are valuable especially for the recent preference learning algorithms and recommender systems. In this paper, we describe the process of mapping and augmenting the movie ratings dataset MovieTweetings from the perspective of RecSysRules 2015 Challenge. The ad-hoc queries to DBpedia are used as an underlying concept. To the best of our knowledge, there is no existing mapping dataset of movies for MovieTweetings. We also provide a brief discussion about the benefits of the augmented feature set for an elementary rule-based representation of the user preferences.

Stať ve sborníku

10.1007/978-3-319-24027-5_11

Katedra softwarového inženýrství

In this paper, we present experiments evaluating Association Rule Classification algorithms on on-line and off-line recommender tasks of the CLEF NewsReel 2014 Challenge. The second focus of the experimental evaluation is to investigate possible performance optimizations of the Classification Based on Associations algorithm. Our findings indicate that pruning steps in CBA reduce the number of association rules substantially while not affecting accuracy. Using only part of the data employed for the rule learning phase in the pruning phase may also reduce training time while not affecting accuracy significantly.

Stať ve sborníku

Katedra softwarového inženýrství

EasyMiner is a web-based visual interface for association rule learning. This paper presents a preview of the next release, which uses the R environment as the data processing backend. EasyMiner/R uses the arules package to learn rules. It uses the Classifications Based on Associations (CBA) algorithm as a classifier and to perform rule pruning. Experimental results show that EasyMiner with the R-based backend is able to handle larger datasets than the previous version.

Stať ve sborníku

10.5220/0005428403900401

Katedra softwarového inženýrství

When a link is not explicitly present in an RDF dataset, it does not mean that the link could not exist in reality. Link prediction methods try to overcome this problem by finding new links in the dataset with support of a background knowledge about the already existing links in the dataset. In dynamic environments that change often and evolve over time, link prediction methods should also take into account the temporal aspects of data. In this paper, we present a novel time-aware link prediction method. We model RDF data as a tensor and take into account the time when RDF data was created. We use an ageing function to model a retention of the information over the time; lower the significance of the older information and promote more recent. Our evaluation shows that the proposed method improves quality of predictions when compared with methods that do not consider the time information.

Stať ve sborníku

Katedra softwarového inženýrství

Client-side execution of a recommender system requires enrichment of the content delivered to the user with a list of potentially related content. A possible bottleneck for client-side recommendation is the data volume entailed by transferring the feature set describing each content item to the client, and the computational resources needed to process this feature set. This paper investigates whether the representation of the textual content (e.g. of videos) with Bag of Entities (BoE) vector generated by a wikifier can yield a classifier with the same accuracy at smaller size than the standard BoW approach. Experimental evaluation performed on the Reuters-21578 text categorization collection shows that there is a small improvement for small term vector sizes.

Stať ve sborníku

Katedra softwarového inženýrství

Tento příspěvek představuje online recommender InBeat.eu. Systém umožňuje sběr explicitních a implicitních zpětných vazeb od uživatelů, které jsou použity jako ukazatele zájmu. Demonstrace systému je zaměřena na interakci uživatelů s multimediálním obsahem, konkrétně se jedná o videa a scénář televizních zpráv. Videa jsou automaticky sémanticky anotována s pomocí nástroje pro hledání pojmenovaných entit. Důležitou součástí sytému je propojení se senzorem Microsoft Kinect, který umožňuje analyzovat natočení hlavy za účelem reálného vyhodnocení sledování daného videa. Ze získaných dat jsou odvozena asociační pravidla představující preference daného uživatele. Tyto pravidla jsou následně použita pro doporučování.

Stať ve sborníku

Katedra softwarového inženýrství

Interest Beat (inbeat.eu) is a service for recommendation of content. InBeat was designed with emphasis on versatility, scalability and extensibility. The core contains the General Analytics INterceptor module, which collects and aggregates user interactions, the Preference Learning module and the Recommender System module. In this paper, we describe InBeat general architecture, putting emphasis on its high- performance architecture that was used in the CLEF-NEWSREEL: News Recommendation Evaluation Lab.

Stať ve sborníku

10.1007/978-3-642-55143-7_8

Katedra softwarového inženýrství

Is it possible to determine only by observing the behavior of a user what are his interests for a media? The aim of this project is to develop an application that can detect whether or not a user is viewing a content on the TV and use this information to build the user profile and to make it evolve dynamically. Our approach is based on the use of a 3D sensor to study the movements of a user’s head to make an implicit analysis of his behavior. This behavior is synchronized with the TV content (media fragments) and other user interactions (clicks, gestural interaction) to further infer viewer’s interest. Our approach is tested during an experiment simulating the attention changes of a user in a scenario involving second screen (tablet) interaction, a behavior that has become common for spectators and a typical source of attention switches.

Stať ve sborníku

10.1007/978-3-319-09870-8_18

Katedra softwarového inženýrství

The main obstacles for a straightforward use of association rules as candidate business rules are the excessive number of rules discovered even on small datasets, and the fact that contradicting rules are generated. This paper shows that Association Rule Classification algorithms, such as CBA, solve both these problems, and provides a practical guide on using discovered rules in the Drools BRMS and on setting the ARC parameters. Experiments performed with modified CBA on several UCI datasets indicate that data coverage rule pruning keeps the number of rules manageable, while not adversely impacting the accuracy. The best results in terms of overall accuracy are obtained using minimum support and confidence thresholds. Disjunction between attribute values seem to provide a desirable balance between accuracy and rule count, while negated literals have not been found beneficial.

Stať ve sborníku

10.3233/978-1-61499-419-0-1227

Katedra softwarového inženýrství

This paper demonstrates Interest Beat (InBeat.eu) as a recommender system for online videos, which determines user interest in the content based on gaze tracking with Microsoft Kinect in addition to explicit user feedback. Content of the videos is represented using a semantic wikifier. User profile is constructed from preference rules, which are discovered with an association rule learner.

Kapitola v knize

Katedra softwarového inženýrství

The rise of new paradigms in the field of television and digital media distribution (e.g. Smart TV, IPTV, Social TV) has opened a new digital world of data communication opportunities but at the same time exacerbated the information overload problem for media consumers and providers. Therefore, the need for personalized content delivery has extended from the traditional web to the networked media domain. This chapter presents a comprehensive research in the field of capturing and representing user preferences and context and an overview of relevant digital media-specific personalized recommendation techniques. Subsequently, it describes the vision and first personalization approach adopted within the LinkedTV EU project, for profiling and contextualizing users and providing targeted information and content in a linked media environment.

Stať ve sborníku

10.1145/2507157.2508217

Katedra softwarového inženýrství

GAIN (inbeat.eu) is a web application and service for capturing and preprocessing user interactions with semantically described content. GAIN outputs a set of instances in tabular form suitable for further processing with generic machine-learning algorithms. GAIN is demoed as a component of a "SMART-TV" recommender system. Content is automatically described with DBpedia types using a Named Entity Recognition (NER) system. Interest is determined based on explicit user actions and user's attention computed by 3D head pose estimation. Preference rules are learnt with an association rule mining algorithm. These can be e.g. deployed to a business rules system, acting as a recommender.

Stať ve sborníku

Katedra softwarového inženýrství

The trend in application development is to provide a personalized interface. The availability of the user preference level associated with user actions is the key for the personalization process. This paper describes a "work-in-progress" framework for deriving user preference from actions performed on semantically annotated objects - be it web pages or TV news. Preference level is computed using supervised learning with genetic programming from implicit feedback, which might be time on page for the web domain, or the user engagement level for the TV domain. We provide tool called GAIN (General Analytics INterceptor) covering the whole approach at wa.vse.cz.

Stať ve sborníku

10.1007/978-3-642-35176-1_3

Katedra softwarového inženýrství

Modelling and understanding various contexts of users is important to enable personalised selection of Web APIs in directories such as Programmable Web. Currently, relationships between users and Web APIs are not clearly understood and utilized by existing selection approaches. In this paper, we present a semantic model of a Web API directory graph that captures relationships such as Web APIs, mashups, developers, and categories. We describe a novel configurable graph-based method for selection of Web APIs with personalised and temporal aspects. The method allows users to get more control over their preferences and recommended Web APIs while they can exploit information about their social links and preferences. We evaluate the method on a real-world dataset from ProgrammableWeb.com, and show that it provides more contextualised results than currently available popularity-based rankings.

Článek

Katedra softwarového inženýrství

Web usage profile is very important in recommender systems. More interesting is the semantic enriched profile, which can describe visitor intents by ontologies and express more information and relations of visitor's character. Our research is based on processing semantically enriched clickstream and application of scoring algorithm, which is based on symbolic regression. A semantic enrichment uses Linked Data principles. The scoring assigns to each pageview a value, which represents and involves visitor interests. Scoring involves all know attributes of each pageview including semantic annotation. The score of each pageview is used to establish a visitor profile. The established profile can be in form of ontologies. In this paper, we propose integrate scoring algorithm into semantic web usage mining and publish visitor profile in RDF/OWL representation. We suggest merge the profiles from different web sites and integrate additional related information from publicly available reso