Online advertising has a continuously increasing popularity. Target audience of this new advertising method is huge. Additionally, there is another rapidly growing and crowded group related to internet advertising that consists of web publishers. Contextual advertising systems make it easier for publishers to present online ads on their web sites, since these online marketing systems automatically divert ads to web sites with related contents. Web publishers join ad networks and gain revenue by enabling ads to be displayed on their sites. Therefore, the accuracy of automated ad systems in determining ad-context relevance is crucial. In this thesis we construct a method for semantic classification of web site contexts in Turkish language and develop an ad serving system to display context related ads on web documents. The classification method uses both semantic and statistical techniques. The method is supervised, and therefore, needs processed sample data for learning classification rules. Therefore, we generate a Turkish marketing dataset and use it in our classification approaches. We form successful classification methods using different feature spaces and support vector machine configurations. Our results present a good comparison between these methods.

Thesis (doctorat) -- Universität Ulm, 2005. / Includes bibliographical references (p. 227-248).

This thesis describes the use of NetFlow data in the systems for detection of disruptions or anomalies in computer network traffic. Various methods for network data collection are described, focusing especially on the NetFlow protocol. Further, various methods for anomaly detection  in network traffic are discussed and evaluated, and their advantages as well as disadvantages are listed. Based on this analysis one method is chosen. Further, test data set is analyzed using the method. Algorithm for real-time network traffic anomaly detection is designed based on the analysis outcomes. This method was chosen mainly because it enables detection of anomalies even in an unlabelled network traffic. The last part of the thesis describes implementation of the  algorithm, as well as experiments performed using the resulting  application on real NetFlow data.

Simulation tools have progressed largely and in modern times they are commonly usedby engineers to design and simulate machines or part of machines before building and deploying them in the field. The field of Hardware-in-the-loop (HIL) is gaining significant interest among companies as they strive to enhance product safety and reliability simul-taneously reducing testing costs and accelerated development speed. This study presents the Real Time simulation improvements effectuated to the Steer and Brake by wire system on an underground face drill rig. These improvements in the model are validated with a comparison between simulated environment and real test data from the machine using a cosimulation between Matlab&Simulink with AMESim. At the end, this improved model is prepared to be compatible with an Hardware-in-the-loop application that requires an adequate computational time.

In recent years, researchers have recognized relational tables on the Web as an important source of information. To assist this research we developed the Dresden Web Tables Corpus (DWTC), a collection of about 125 million data tables extracted from the Common Crawl (CC) which contains 3.6 billion web pages and is 266TB in size. As the vast majority of HTML tables are used for layout purposes and only a small share contains genuine tables with different surface forms, accurate table detection is essential for building a large-scale Web table corpus. Furthermore, correctly recognizing the table structure (e.g. horizontal listings, matrices) is important in order to understand the role of each table cell, distinguishing between label and data cells. In this paper, we present an extensive table layout classification that enables us to identify the main layout categories of Web tables with very high precision. We therefore identify and develop a plethora of table features, different feature selection techniques and several classification algorithms. We evaluate the effectiveness of the selected features and compare the performance of various state-of-the-art classification algorithms. Finally, the winning approach is employed to classify millions of tables resulting in the Dresden Web Table Corpus (DWTC).

Epiroc Rock Drills AB in Örebro manufactures and develops machinery for the infrastructure required to maintain a fully operational mining industry, such as ventilation, drilling rigs, trucks and loaders etc. In these environmentally conscious times, a large market focus forzero emission machines has begun to emerge. By replacing today'sinternal combustion engines, mining companies can save large sums of money in ventilation costs and fuel while creating a more pleasant working environment in the mines. Due to these rapidly changing needs, Epiroc's machines must also change in design and performance. Epiroc has thus chosen to use interchangeable batteries in their new generation of mining machines. When the battery is discharged, it must be easy to replace with a fullycharged one. From an economic point of view, it is also important that the machine is always in production without any stops, making fast and efficient battery switches desirable. At present, the battery change is usually done with a mono-rail crane down the mine. This has proved to be very difficult and ineffective as the battery is too often jammed into the machine due to the fact that the machine is poorly positioned against the crane. Ceiling heightis also a problem, preferably one should have about 6-7 meters to accomplish a safe lift with the crane, which is not always possible down in a mine. In order to find a more long-term solution to this problem, this degree project took its start. During the course of the process, a number of methods in product development and concept generation have been used to develop different concepts and screen them based on the needs identified at the beginning of the work. The main problem was divided into four sub problems to facilitate conceptgeneration, the most promising sub concepts were visualized using CAD and then put together in different combinations. These combinations were then evaluated against today's solution to easily see what concept should be further developed. This resulted in a concept that consists of a separate platform,which can either be placed on a vehicle or as a stand-alone station. On this platform there are two telescopic arms, one on each side. These arms lifts the three ton heavy battery from the mining machine in an arcuate motion over a charged battery on the platform, placing it in a designated location and then lifts the charged battery into the machine. With this concept, it is also possible to accomplish a safe change with a limited ceiling height of approximately 3.5-4 meters, which is an improvement compared to today's solution.

(in English): The thesis is devoted to the comparison of a few methods of mining knowledge from data. Methods decision tree, classification rules, cluster analysis, and Naive Bayes classifier were applied to the data sample. Data about clients of a non-profit organization Association of Civil Counseling were used. It has been worked according to the technological process of knowledge mining. In the thesis was applied data description, data preparation, modeling and testing and results from interpretation. Because of using the same sample of data and similar data preparation, overlapping results are also expected. The research is focused not only on results similarity, but also differences in results. The correlation between the amount of debt of clients and other attributes was found. In the results, there really were some patterns repeating through most of all methods. It turned out the amount of debt is related to a number of creditors. The more creditors, the higher amount of debt. Clients with a higher amount of liabilities had also higher debt. The results might not be surprising, but it proves the functionality of models and comparability of results.

Útoky typu odoprenia služby (DDoS) sú v dnešných počítačových sieťach stále frekventovanejším bezpečnostným incidentom. Táto práca sa zameriava na detekciu týchto útokov a poskytnutie relevantných informácii za účelom ich mitigácie v reálnom čase. Spomínaná funkcionalita je dosiahnutá s využitím techník prúdového dolovania z dát a strojového učenia. Výsledkom práce je sada nástrojov zastrešujúca celý proces strojového učenia - od vlastnej extrakcie príznakov cez predspracovanie dát až po export natrénovaného modelu pripraveného na nasadenie v produkcii. Experimentálne výsledky vyhodnotené na viacerých reálnych a syntetických dátových sadách poukazujú na presnosť systému väčšiu ako 99% s možnosťou spoľahlivej detekcie prebiehajúceho útoku do 4 sekúnd od jeho začiatku.

Submitted by SBI Biblioteca Digital (sbi.bibliotecadigital@puc-campinas.edu.br) on 2018-02-21T17:47:30Z No. of bitstreams: 1 JOSE DIVINO DE LIMA.pdf: 3046765 bytes, checksum: a793279094d547961482cafe99be62cb (MD5) / Made available in DSpace on 2018-02-21T17:47:30Z (GMT). No. of bitstreams: 1 JOSE DIVINO DE LIMA.pdf: 3046765 bytes, checksum: a793279094d547961482cafe99be62cb (MD5) Previous issue date: 2017-08-31 / The present dissertation work analyses telecommunication systems failures caused by internal and external agents. This analysis can be very challenging since such systems are complex and heterogeneous. Within this context, this work proposed a model that can be used to predict consequent failures from data samples. To do so, we have used a data mining tool and prediction algorithms that create decision trees. Applying the proposed model to a set of faults, generated by the system of a major telecommunications operator, it was demonstrated that it is possible to group faults with an accuracy of 85.96%. In this way, a process can be established that assists in the definition of grouping and correlation of failures, which allows that high level management systems can be configured more efficiently by their administrators. / O presente trabalho de disserta??o tem como principal objetivo a an?lise dos sistemas de telecomunica??o, os quais est?o cada vez mais complexos e heterog?neos e, em fun??o disso, suscet?veis a diversos tipos de falhas causadas tanto por fatores internos como externos, sendo estes ?ltimos devido ? integra??o com sistemas de terceiros. Dentro desse contexto, este trabalho apresenta, ent?o, um modelo que pode ser utilizado para prever falhas consequentes a partir de uma amostra de dados. Para tanto, utilizou-se uma ferramenta de minera??o de dados e algoritmos de predi??o, que criam ?rvores de decis?o. Aplicado o modelo proposto a um conjunto de falhas, gerado pelo sistema de uma grande operadora de telecomunica??es, demonstrou-se que ? poss?vel agrupar falhas com precis?o de 85,96%. Logo, pode-se estabelecer um processo que auxilia na defini??o do agrupamento e correla??o de falhas, permitindo que os sistemas de gest?o de alto n?vel possam ser configurados de maneira mais eficiente pelos administradores.

In this thesis we try using machine learning techniques to predict final grade of a student in a learning management system on the basis of his behavior during the semester. The aim is to determine the optimal technology for the extraction, treatment and machine learning on data. The whole system would then be implemented as a module that we will be able to plug in the existing system.