Time Series Analysis and Market Microstructure Aspects on Short Time Scales

ARMA-GARCH and FIGARCH models with non-normal, tempered-stable innovations are applied to intraday financial time-series on high-frequency time scales. The goal is to investigate their risk forecasting performance and to observe random scaling behavior. To this end, Value-at-Risk (VaR) and Conditional VaR is predicted. In the second part limit order books based on real trading information are modeled with a statistical-learning approach in order to forecast the execution of passive limit orders.