|
|
Recognition of vehicles using signals sensed by smartphone
Nevěčná, Leona ; Vítek, Martin (referee) ; Smíšek, Radovan (advisor)
Thanks to the development in recent years, the placement of miniaturized sensors such as accelerometers, gyroscopes, magnetometers, global positioning system receivers (GPS), microphones or others to commercially sold smartphones is increasing. Use of these sensors (which are to be found in the smartphone) for human activity recognition with health care improvement in mind is a discussed theme. Advantages of the use of smartphone for human movement monitoring lies in the fact that it is a device that the person measured carries with them and there are no additional costs. The disadvantages are a limited storage and battery. Therefore, only accelerometer, gyroscope, magnetometer, and microphone were chosen because their combination achieves best results. GPS sensor was excluded for its lack of reliability in sampling and for being energy demanding. Features were computed from the measured data and used for learning of the classification model. The highest accuracy was achieved with the use of a machine learning method called Random Forest. The main goal of this work was to create an algorithm for transportation mode recognition using signals sensed by a smartphone. The created algorithm succeeds in classification of walk, car, bus, tram, train, and bike in 97.4 % with 20 % holdout validation. When tested on a new set of data from the tenth volunteer, the resulting accuracy counted as average form classification recall for each transportation mode reached 90.49 %.
|
|
|
Dynamics of the volume-volatility relationship in the currency markets
Tůma, Adam ; Baruník, Jozef (advisor) ; Komárek, Luboš (referee)
This work investigates the volume-volatility relationship dynamics in the currency markets using data of five currency pairs in the period between 2010 and 2022. By employing multiple specifications of the HAR model with volume- related regressors and also with time-varying parameters (TVP), we examine the relationships' changing dynamics over time with a focus on improving volatility forecasting performance. Our main findings suggest a strong correlation between volume and volatility. The TVP-HARV model shows significantly changing dy- namics of the volume-volatility relationship, especially during periods affected by politics, changing monetary policies or global crises. The proposed models, however, do not improve out-of-sample volatility forecasting performance com- pared to the benchmark HAR model. The causal effect in the volume-volatility relationship in the currency markets is slightly more substantial in the direction of volatility towards volume, where we find slight forecasting improvements. Our findings conclude that volume and volatility in the currency markets are mainly moving simultaneously with a very strong correlation and much weaker and often insignificant causal effects on both sides, which supports the mixture of distributions hypothesis.
|
|
|
Volume - volatility relation across different volatility estimators
Kvasnička, Tomáš ; Krištoufek, Ladislav (advisor) ; Avdulaj, Krenar (referee)
The main objective of this thesis is to analyze whether traded volume increases predictive power of volatility. We are mostly focused on Garman-Klass volatility estimator, which is more efficient than squared returns. Both univariate (AR, HAR, ARFIMA) and multivariate models (VAR, VAR-HAR) are used to find out if traded volume improves volatility forecasting. Furthermore, GARCH(1,1) both with and without traded volume is carried out and forecasted. All these methods are estimated on a basis of rolling window and during each step 1-day ahead forecast is computed. Final assessment is based on MAPE, RMSE and Mincer-Zarnowitz test of the out-of-sample forecasts, which are compared with the realized volatility. It turns out that traded volume slightly improves predictive power of the scrutinized models in case of FTSE 100 and IPC Mexico, contrary to Nikkei 225 and S&P 500 when a decrease of the predictive power is detected. Moreover, we observe that only HAR and VAR-HAR models are able to produce an unbiased forecast. As the evidence of the improvement is not conclusive and to maintain model parsimony, HAR model fitted by Garman-Klass volatility appears to be the best alternative in case of missing the realized volatility.
|
|
|
Recognition of vehicles using signals sensed by smartphone
Nevěčná, Leona ; Vítek, Martin (referee) ; Smíšek, Radovan (advisor)
Thanks to the development in recent years, the placement of miniaturized sensors such as accelerometers, gyroscopes, magnetometers, global positioning system receivers (GPS), microphones or others to commercially sold smartphones is increasing. Use of these sensors (which are to be found in the smartphone) for human activity recognition with health care improvement in mind is a discussed theme. Advantages of the use of smartphone for human movement monitoring lies in the fact that it is a device that the person measured carries with them and there are no additional costs. The disadvantages are a limited storage and battery. Therefore, only accelerometer, gyroscope, magnetometer, and microphone were chosen because their combination achieves best results. GPS sensor was excluded for its lack of reliability in sampling and for being energy demanding. Features were computed from the measured data and used for learning of the classification model. The highest accuracy was achieved with the use of a machine learning method called Random Forest. The main goal of this work was to create an algorithm for transportation mode recognition using signals sensed by a smartphone. The created algorithm succeeds in classification of walk, car, bus, tram, train, and bike in 97.4 % with 20 % holdout validation. When tested on a new set of data from the tenth volunteer, the resulting accuracy counted as average form classification recall for each transportation mode reached 90.49 %.
|
|
|
Understanding co-jumps in financial markets
Thoma, Richard ; Baruník, Jozef (advisor) ; Vošvrda, Miloslav (referee)
This thesis focuses on impact of jumps and simultaneous jumps (co-jumps) in asset prices on future volatility. Our main contribution to the empirical literature lies in the use of panel Heterogeneous Autoregressive (HAR) model that allows us to obtain average effect of jumps for both the portfolio of 29 U.S. stocks and 8 individual market sectors our stocks belong to. On top of that we investigate the effect of sign for both jumps and co-jumps. The estimation results indicate that the impact of jumps on future volatility is positive whereas for co-jumps it is negative. We also document tendency of downward jumps and co-jumps to be followed by increase in volatility and that upward jumps and co-jumps are followed by decrease in volatility. Finally, results for individual sectors reveal that estimated effects vary across industries - for cyclical sectors volatility is in general more sensitive to negative jumps and less sensitive to positive jumps than for defensive sectors.
|
|
|
Realized Jump GARCH model: Can decomposition of volatility improve its forecasting?
Poláček, Jiří ; Baruník, Jozef (advisor) ; Pertold-Gebicka, Barbara (referee)
The present thesis focuses on exploration of the applicability of realized measures in volatility modeling and forecasting. We provide a first comprehensive study of jump variation impact on future volatility of Central and Eastern European stock markets. As a main workhorse, the recently proposed Realized Jump GARCH model, which enables a study of the impact of jump variation on future volatility forecasts, is used. In addition, we estimate Realized GARCH and heterogeneous autoregressive (HAR) models using one-minute and five-minute high frequency data. We find that jumps are important for future volatility, but only to a limited extent due to the high level of information aggregation within the stock market index. Moreover, Realized (Jump) GARCH models outperform the standard GARCH model in terms of data fit and forecasting performance. Comparison of forecasts with HAR models reveals that Realized (Jump) GARCH models capture higher portion of volatility variation. Eventually, Realized Jump GARCH compared to other Realized GARCH models provides comparable or even better forecasting performance.
|
|
|
Realized Jump GARCH model: Can decomposition of volatility improve its forecasting?
Poláček, Jiří ; Baruník, Jozef (advisor) ; Pertold-Gebicka, Barbara (referee)
The present thesis focuses on exploration of the applicability of realized measures in volatility modeling and forecasting. We provide a first comprehensive study of jump variation impact on future volatility of Central and Eastern European stock markets. As a main workhorse, the recently proposed Realized Jump GARCH model, which enables a study of the impact of jump variation on future volatility forecasts, is used. In addition, we estimate Realized GARCH and heterogeneous autoregressive (HAR) models using one-minute and five-minute high frequency data. We find that jumps are important for future volatility, but only to a limited extent due to the high level of information aggregation within the stock market index. Moreover, Realized (Jump) GARCH models outperform the standard GARCH model in terms of data fit and forecasting performance. Comparison of forecasts with HAR models reveals that Realized (Jump) GARCH models capture higher portion of volatility variation. Eventually, Realized Jump GARCH compared to other Realized GARCH models provides comparable or even better forecasting performance.
|
|
|
Volume - volatility relation across different volatility estimators
Kvasnička, Tomáš ; Krištoufek, Ladislav (advisor) ; Avdulaj, Krenar (referee)
The main objective of this thesis is to analyze whether traded volume increases predictive power of volatility. We are mostly focused on Garman-Klass volatility estimator, which is more efficient than squared returns. Both univariate (AR, HAR, ARFIMA) and multivariate models (VAR, VAR-HAR) are used to find out if traded volume improves volatility forecasting. Furthermore, GARCH(1,1) both with and without traded volume is carried out and forecasted. All these methods are estimated on a basis of rolling window and during each step 1-day ahead forecast is computed. Final assessment is based on MAPE, RMSE and Mincer-Zarnowitz test of the out-of-sample forecasts, which are compared with the realized volatility. It turns out that traded volume slightly improves predictive power of the scrutinized models in case of FTSE 100 and IPC Mexico, contrary to Nikkei 225 and S&P 500 when a decrease of the predictive power is detected. Moreover, we observe that only HAR and VAR-HAR models are able to produce an unbiased forecast. As the evidence of the improvement is not conclusive and to maintain model parsimony, HAR model fitted by Garman-Klass volatility appears to be the best alternative in case of missing the realized volatility.
|
|
|
Portfólio Value at Risk a Expected Shortfall s použitím vysoko frekvenčních dat
Zváč, Marek ; Fičura, Milan (advisor) ; Janda, Karel (referee)
The main objective of this thesis is to investigate whether multivariate models using Highfrequency data provide significantly more accurate forecasts of Value at Risk and Expected Shortfall than multivariate models using only daily data. Our objective is very topical since the Basel Committee announced in 2013 that is going to change the risk measure used for calculation of capital requirement from Value at Risk to Expected Shortfall. The further improvement of accuracy of both risk measures can be also achieved by incorporation of high-frequency data that are rapidly more available due to significant technological progress. Therefore, we employed parsimonious Heterogeneous Autoregression and its asymmetric version that uses high-frequency data for the modeling of realized covariance matrix. The benchmark models are chosen well established DCC-GARCH and EWMA. The computation of Value at Risk (VaR) and Expected Shortfall (ES) is done through parametric, semi-parametric and Monte Carlo simulations. The loss distributions are represented by multivariate Gaussian, Student t, multivariate distributions simulated by Copula functions and multivariate filtered historical simulations. There are used univariate loss distributions: Generalized Pareto Distribution from EVT, empirical and standard parametric distributions. The main finding is that Heterogeneous Autoregression model using high-frequency data delivered superior or at least the same accuracy of forecasts of VaR to benchmark models based on daily data. Finally, the backtesting of ES remains still very challenging and applied Test I. and II. did not provide credible validation of the forecasts.
|
guest ::
