Financial institutions have a tough job managing and constantly watching the risk in many portfolios and they deal with a lot of data. To make this easier an automatic system that can spot unusual patterns in portfolio risk measures would be very useful. Therefore, this study looked at four methods for detecting these unusual patterns in time series data the Autoregressive Integrated Moving Average (ARIMA)-Generalized Autoregressive Conditional Heteroscedasticity (GARCH), and Exponentially Weighted Moving Average (EWMA), Long Short-Term Memory (LSTM) and Hierarchical Temporal Memory (HTM) Machine Learning (ML) algorithms. We tested these methods using three sets of synthetic data and one set of real-world data with manually added labels. The findings exhibit that LSTM networks are effective at this task. On the other hand, EWMA models are faster and easier to understand. The ARMA-GARCH model didn't fit the time series data well and performed poorly. The HTM method was outperformed by the other methods because it struggled to learn the time series patterns needed for effective detection. In short, LSTM models are the best at finding anomalies in portfolio risk measures. EWMA models are good for speed and clarity, but ARMA-GARCH and HTM models did not perform as well.