1,720,958 research outputs found
Application of decision trees to failure detection in HVAC systems
Full text linkDiplomová práce se zabývá využitím rozhodovacích stromů při návrhu pravidel pro detekci poruch technických zařízení budov. V první části této práce jsou představeny koncepty strojového učení a jsou popsány rozhodovací stromy. Na základě teoretických poznatků jsou vybrány vhodné metody rozhodovacích stromů a to CART, Random Forest a šikmé rozhodovací stromy (OC1). V dalších sekcích je představena implementace vybraných metod a jsou popsány doporučené postupy, jak dané metody používat a nastavit jejich parametry. V poslední části diplomové práce jsou vybrané metody porovnány na sadách reálných dat z technických zařízení budov. Jednodušší část testování proběhla na otopných okruzích a následně byly metody použity na diagnostiku vzduchotechnický jednotek.This thesis seeks to analyze the usage of decision trees in diagnostics rule design in HVAC. At first, theoretical concepts of machine learning and decision trees are introduced. Viable methods are selected based on theoretical research. These methods are CART, Random Forest and oblique decision tree classifier (OC1). Implementation of selected methods is described as well as their limitations and guide to parameter tuning. Lastly, methods are compared using real-world datasets from HVAC systems. The comparison starts at more straightforward tasks from heating circuits and then follow up with problems from the air handling units domain
Game Theoretic Optimization of Detecting Malicious Behavior
No full textKlasifikátory založené na strojovém učení se používají v takových bezpečnostních aplikacích jako detekce podvodného chování nebo detekce narušení bezpečnosti počítačových ve snaze předejít detekci. Avšak klasické metody strojového učení se s tím nedokážou vypořádat jelikož jsou založené na předpokladu že budoucí pozorování budou odpovídat rozložení trenovácích dat. S použitím relevantní literatury možné útoky na klasifikátory vyplývající z této jejich limitace jsou analyzovány spolu s existujícími metodami reakce na nebezpečí útoků. Diskutujeme že místo ignorování existence adaptability útočníků a opravy následně způsobené škody je výhodnější namodelovat soupeře pomocí teorie her s následnou predikcí a omezením jeho možností způsobovat škodu. Jenže byla zjištěna mezera mezi praktickými požadavky na klasifikátory a vlastnostmi existujících modelů herně teoretické optimalizace detekce škodlivého chování. V této práci vyvíjíme postup který vyplní danou mezeru. Praktická aplikovatelnost navržené metody byla vynucena spoluprací s Divizí Bezpečnosti společnosti O2 Czech Republic a.s.: nová metodabyla vyvinuta jako vylepšení interního systému na detekci podvodného chování. Ve výsledku navrženou metodu se dá aplikovat na binární klasifikátor jako na černou skříňku, bez omezení na použitý algoritmus strojového učení. Navíc vyvinutý postup umožňuje omezovat poměr falešných poplachů, což je zásadním požadavkem v bezpečnostních aplikacích strojového učení. Dále model bere v úvahu skutečnost že existují různé typy útočníka. Kromě toho variabilita typů útočníka a zisky v rámci modelu jsou odvozený na základě nasbíraných datech, což minimalizuje počet hypotéz nepodložených pozorováními. Taky model bere v potaz omezenou racionalitu útočníků. Součástí pozorováními. Taky model bere v potaz omezenou racionalitu útočníků. Součástí postupu jsou i vyvinuté efektivní algoritmy na počítání několika herně teoretických konceptů řešení: Nashovy rovnováhy, Stackelbergovy rovnováhy i Stackelbergovy rovnováhy za omezení na poměr falešných poplachů. Díky efektivitě navržených algoritmů je očekáváno že vyvinutý postup zůstane aplikovatelný i pro větší soubory dat. Nakonec efektivita postupu je demonstrována pomocí rozsáhlé experimentální evaluace. S využitím dat z klasifikátorů pro detekce narušení bezpečnosti bylo ukázáno že navržené algoritmy jsou lepší než existující alternativy. Dále na případě detekce podvodného chování v O2 Czech Republic je ukázáno že vyvinutá metoda zachovává účinnost klasifikátoru bez herně teoretické optimalizace na statických datech a vylepšuje robustnost klasifikace pokud se útočník chová v souladu s navrženým modelem.Machine learning classifiers are used in security applications such as fraud detection or intrusion detection in computer networks. In applications of this kind the classified entities tend to evolve in time attempting to avoid detection. However, classical machine learning methods fail to address the issue, assuming that future observations would follow the same distribution as training data. Based on related work we analyze possible attacks against a classifier arising due to this limitation and survey existing approaches to deal with the attacks. We discuss that rather than ignoring the adaptivity and repairing the damage once it occurs, it is more advantageous to model the adversary by means of game theory and mitigate his ability to cause the damage in a predictive manner. Yet, we identify a gap between practical requirements on adversarial classifiers and properties of the present methods for game theoretic optimization of detecting malicious behavior. In this thesis we develop an approach filling the gap. Practical applicability of the method was enforced by the collaboration with the Security Division of O2 Czech Republic telecommunications company: the novel method was developed as an improvement for the company?s internal fraud detection system. As a result, the devised method can be applied to any binary classifier as a black box, not limiting the modeling power of the used machine learning algorithm. Moreover, the approach enables restricting a false alarm rate, satisfying a crucial requirement in the security domain. Furthermore, the model takes into consideration the fact that there are different types of adversaries. In addition, both variability of the adversary types and utilities in the model are derived based on collected data, minimizing hypotheses unfounded with observations. The model also addresses the bounded rationality of adversaries. As part of the approach, we develop efficient algorithms for computation of several solution concepts: Nash equilibrium, Stackelberg equilibrium and Stackelberg equilibrium under the restriction on false alarm rate. Thanks to the algorithms efficiency the approach is expected to remain applicable in case of large datasets. Finally, the efficacy of the developed approach is demonstrated via extensive experimental evaluation. Using data from real-world intrusion detection classifiers, it is shown that the developed algorithms are superior compared to available alternatives. Next, on the case of O2 Czech Republic fraud detection it is demonstrated that the developed method preserves out-of-sample performance of the classifier without the game theoretic optimization, while improving robustness of the classification when the attacker behaves in accordance with the model
Computing and evaluating pricing strategies in price comparison shopping
No full textkatedra řídicí technik
Going Beyond Counting First Authors in Author Co-citation Analysis
Full text linkThe present study examines one of the fundamental aspects of author co-citation analysis (ACA) - the way co-citation
counts are defined. Co-citation counting provides the data on which all subsequent statistical analyses and mappings
are based, and we compare ACA results based on two different types of co-citation counting - the traditional type that
only counts the first one among a cited work's authors on the one hand and a non-traditional type that takes into
account the first 5 authors of a cited work on the other hand. Results indicate that the picture produced through this non-traditional author co-citation counting contains more coherent author groups and is therefore considerably clearer. However, this picture represents fewer specialties in the research field being studied than that produced through the traditional first-author co-citation counting when the same number of top-ranked authors is selected and analyzed. Reasons for these effects are discussed
Variations on the Author
Full text link“Variations on the Author” discusses two of Eduardo Coutinho’s recent films (Um Dia na Vida, from 2010, and Últimas Conversas, posthumously released in 2015) and their contribution to the general question of documentary authorship. The director’s filmography is characterized by a consistent yet self-effacing form of authorial self-inscription: Coutinho often features as an interviewer that rather than express opinions propels discourses; an interviewer that is good at listening. This mode of self-inscription characterizes him as an author who is not expressive but who is nonetheless markedly present on the screen. In Um Dia na Vida, however, Coutinho is completely absent form the image, while Últimas Conversas, on the contrary, includes a confessional prologue that moves the director from the margins to the center of his films. This article examines the ways in which these works stand out in the filmography of a director who offers new insights into the notion of cinematic authorship
Appropriate Similarity Measures for Author Cocitation Analysis
Full text linkWe provide a number of new insights into the methodological discussion about author cocitation analysis. We first argue that the use of the Pearson correlation for measuring the similarity between authors’ cocitation profiles is not very satisfactory. We then discuss what kind of similarity measures may be used as an alternative to the Pearson correlation. We consider three similarity measures in particular. One is the well-known cosine. The other two similarity measures have not been used before in the bibliometric literature. Finally, we show by means of an example that our findings have a high practical relevance.information science;Pearson correlation;cosine;similarity measure;author cocitation analysis
Dispelling the Myths Behind First-author Citation Counts
Full text linkWe conducted a full-scale evaluative citation analysis study of scholars in the XML research field to explore just how different from each other author rankings resulting from different citation counting methods actually are, and to demonstrate the capability of emerging data and tools on the Web in supporting more realistic citation counting methods. Our results contest some common arguments for the continued
use of first-author citation counts in the evaluation of scholars, such as high correlations between author rankings by first-author citation counts and other citation
counting methods, and high costs of using more realistic citation counting methods that are not well-supported by the ISI databases. It is argued that increasingly available digital full text research papers make it possible for citation analysis studies to go beyond what the ISI databases have directly supported and to employ more
sophisticated methods
koamabayili/VECTRON-author-checklist: VECTRON author checklist
No full textWe have done our best to complete the author checklist relating to the use of animals in the hut study. Note that the objective for the hut study was to evaluate the IRS treatment applications for residual efficacy against Anopheles mosquitoes, including the local An. coluzzii mosquito population. Cows were only used to attract mosquitoes into the huts and no tests were carried out directly on the cows. The author checklist is intended for use with studies where experiments are carried out on animals, which is why we have had such difficulty in completing this for the hut study, as many of the questions do not relate to how the cows were used
- …
