|
|
Mispricing in leveraged value small-capitalization stocks
Picálek, Jan ; Hronec, Martin (advisor) ; Novák, Jiří (referee)
We study returns in the universe of leveraged value small-capitalization stocks, a universe with historically significant exposure to common risk factors. We sep- arate future winners and losers within this universe of risky stocks by adopting machine-learning-based mispricing strategy. The strategy considers 34 stock- level characteristics to predict 1-month-ahead returns and construct a long- short portfolio accordingly. The portfolio yields abnormal risk-adjusted re- turns of 0.42% per month out-of-sample, uncovering statistically significant mispricing. The machine-learning algorithm is trained on leveraged value small- capitalization stocks, so it captures universe-specific nonlinearities and variable interactions. The nonlinear effects and predictive power of individual variables are extracted and presented as well. We found no evidence of a relationship between the magnitude of the mispricing and credit cycles, or market volatility. JEL Classification G11, G12, G14, Keywords Anomalies, Predictability of returns, Asset pricing tests, Leveraged equities, Value stocks Title Mispricing in leveraged value small-capitalization stocks
|
|
|
Detection of Anomalies in Pedestrian Walking
Pokorný, Ondřej ; Orság, Filip (referee) ; Goldmann, Tomáš (advisor)
The goal of this work was to create a system that would be able to detect anomalies in pedestrian walking. As the core of my application, I have used OpenPose, which is an application for detecting human skeletons. Then I used a bidirectional LSTM neural network to detect anomalies in video sequences. This architecture was chosen during the experiment because it outperformed other solutions. I trained my model to detect three types of anomalies. The output of my application is a video with marked sequences of anomalies. The whole system is implemented in Python.
|
|
|
Browser and User Fingerprinting for Practical Deployment
Vondráček, Tomáš ; Malinka, Kamil (referee) ; Polčák, Libor (advisor)
The aim of the diploma thesis is to map the information provided by web browsers, which can be used in practice to identify users on websites. The work focuses on obtaining and subsequent analysis of information about devices, browsers and side effects caused by web extensions that mask the identity of users. The acquisition of information is realized by a designed and implemented library in the TypeScript language, which was deployed on 4 commercial websites. The analysis of the obtained information is carried out after a month of operation of the library and focuses on the degree of information obtained, the speed of obtaining information and the stability of information. The dataset shows that up to 94 % of potentially different users have a unique combination of information. The main contribution of this work lies in the created library, design of new methods of obtaining information, optimization of existing methods and the determination of quality and poor quality information based on their level of information, speed of acquisition and stability over time.
|
|
|
Detection of DoS and DDoS attacks targeting a web server
Nguyen, Minh Hien ; Fujdiak, Radek (referee) ; Kuchař, Karel (advisor)
The bachelor thesis deals with the detection of DoS (Denial of service) and DDoS (Distributed Denial of Service) attacks targeting a web server. This work aims to design detection methods, which will be subsequently tested. Analysis of attacks according to the ISO/OSI (International Organization for Standardization/Open Systems Interconnection) reference model will allow an understanding of the features of individual attacks. In the practical part, some tools are used to implement attacks, then there are generators of legitimate network traffic and a secure web server. Substantial data are created from ongoing attacks and communications of ordinary users. These data are an important part of the proposed methods. The purpose of assessing the achieved results is to evaluate the effectiveness of individual detection methods in terms of accuracy and time consumption.
|
|
|
Multi-horizon equity returns predictability via machine learning
Nechvátalová, Lenka ; Baruník, Jozef (advisor) ; Krištoufek, Ladislav (referee)
We examine the predictability of expected stock returns across horizons using machine learning. We use neural networks, and gradient boosted regression trees on the U.S. and international equity datasets. We find that predictabil- ity of returns using neural networks models decreases with longer forecasting horizon. We also document the profitability of long-short portfolios, which were created using predictions of cumulative returns at various horizons, be- fore and after accounting for transaction costs. There is a trade-off between higher transaction costs connected to frequent rebalancing and greater returns on shorter horizons. However, we show that increasing the forecasting hori- zon while matching the rebalancing period increases risk-adjusted returns after transaction cost for the U.S. We combine predictions of expected returns at multiple horizons using double-sorting and buy/hold spread, a turnover reduc- ing strategy. Using double sorts significantly increases profitability on the U.S. sample. Buy/hold spread portfolios have better risk-adjusted profitability in the U.S. JEL Classification G11, G12, G15, C55 Keywords Machine learning, asset pricing, horizon pre- dictability, anomalies Title Multi-horizon equity returns predictability via machine learning
|
|
|
Specific anomaly detection methods in wireless communication networks
Holasová, Eva ; Blažek, Petr (referee) ; Fujdiak, Radek (advisor)
The diploma thesis is focuses on technologies and security of the wireless networks in standard IEEE 802.11, describes the most used standards, definition of physical layer, MAC layer and specific technologies for wireless networks. The diploma thesis is focused on description of selected security protocols, their technologies as well as weaknesses. Also, in the thesis, there are described security threats and vectors of attacks towards wireless networks 802.11. Selected threats were simulated in established experimental network, for these threats were designed detection methods. For testing and implementing designed detection methods, IDS system Zeek is used together with network scripts written in programming language Python. In the end there were trained and tested models of machine learning both supervised and unsupervised machine learning.
|
|
|
Network Traffic Analysis Based on Sketches
Dřevo, Aleš ; Kekely, Lukáš (referee) ; Bartoš, Václav (advisor)
Aim of this thesis is to create a program for network traffic analysis and for detection of anomallies in the traffic. The Heavy-Changes Detection technique which falls within the Data stream algorithm category is used to do so. Special structures called sketches are used for data processing. These structures are capable of maintaining large amounts of data with low memory consumption. Programs from Nemea system for which this project is created are used for gathering necessary network data.
|
|
| |
|
|
Anomaly detection by neural networks
Strakoš, Jan ; Sikora, Marek (referee) ; Blažek, Petr (advisor)
This bachelor thesis is focused on anomaly detection represented as computer network attacks by neural network. One of the most common groups of attacks is Distributed Denial of Service (DDoS) attacks, which the system based on neural network should identificate. In the theoretical part of this thesis are described legitimate, non-standard and illegitimate traffic. Another part of this chapter described DDoS attacks, options of their detection, neural networks principle and their use. Practical part describe choosed communication parameters, specifying the threshold intervals of legitimate traffic, constructing a neural network which use of these parameters and threshold intervals, implementation of neural network into the system and presenting results.
|
|
|
Parametrization of network attacks
Jelínek, Michael ; Sikora, Pavel (referee) ; Blažek, Petr (advisor)
This bachelor thesis is dedicated to the definition of suitable parameters for network attack identification with the use of neural networks. In the theoretical part of this thesis are methods for anomaly detection in network communication, structure of artificial neural networks and DDoS attacks used for verification of detection capabilities. The practical part of this thesis is focused on the process of preparing data, the subsequent implementation into neural networks and a summary of the results achieved for the different setups of neural networks.
|
guest ::
