Advances in modern medicine are constantly improving clinical patient safety. Nevertheless, the perioperative environment remains a high-risk setting for patients. A major threat connected to this period is perioperative organ injury, including mainly cardiovascular complications, which are often only detected at a later stage. Currently, the standard procedure for patients risk evaluation during the perioperative period, includes preoperative evaluation (e.g. imaging, laboratory tests, ECG) combined with assessment of risk classification systems (Revised Cardiac Risk Index (RCRI), American Society of Anesthesiologists (ASA) score). These scores are highly subjective and only capture the general risk of the patients. For better and continuous quantification of the patients risk, an on site risk monitoring system would help to detect possible upcoming complications at an earlier stage, which is not possible with current measurement systems. There is evidence that a compromised autonomic nervous system (ANS) has a strong correlation with adverse cardiovascular events. Therefore, the assessment of the ANS activity in the perioperative period might be used as a risk indicator. A possible method to observe the ANS activity is heart rate variability (HRV) analysis. To strengthen the rational for this analysis, this work aims to provide insights in the perioperative HRV behaviour. For this purpose, ECG of 85 patients undergoing general surgery was measured. The HRV was then extracted for each patient and the course of the signals during surgery and post-anesthesia recovery analysed. Using statistical analysis, HRV features, representative for the risk classes, were selected as input for different classification algorithms and their performance was evaluated. Results show a general course of the HRV in all risk groups, with a strong decrease during anesthesia followed by an increase in the recovery period. Moreover, patients in lower risk groups show a stronger increase of the HRV values during recovery, suggesting a higher autonomic activity. The classification algorithms performed only mediocre with the chosen input HRV features, reaching a top accuracy of F1=0.82 (0.28) for GRA classification. Concluding, it can be said that perioperative HRV does show risk related patterns, however the used HRV features and models are not suitable to classify patients accurately enough for clinical use, indicating the need of further research. In a future work, more patients could be analysed, different featuresets calculated or different classification models implemented. In the course of this work two conference papers were submitted.