PCA and Neural Networks for Financial Time Series Anomaly Detection

Summary

This approach detects anomalies in financial time series containing multiple market risk factors, where contaminated observations can distort risk models and their resulting measures. It first applies principal component analysis to compress and reconstruct time-series data, extracting features that represent the series. A feedforward neural network then produces an anomaly score, and observations above a cutoff are flagged.

Rather than setting that cutoff manually, the method learns it as a network parameter through a customized loss function. The authors compare the approach with established anomaly detectors on synthetic and real datasets, reporting high and stable performance. They also find lower value-at-risk estimation errors when detected anomalies are corrected using a basic imputation method. The excerpt does not specify dataset sizes, comparison settings, or how well the results hold across markets and regimes; imputation can also affect downstream risk estimates.

Key ideas

• Principal component analysis compresses and reconstructs financial series to extract detection features.
• A feedforward neural network assigns an anomaly score to each series.
• The anomaly cutoff is learned during optimization rather than fixed by hand.
• The approach is evaluated against known detectors on synthetic and real data.
• Basic imputation after detection is reported to reduce value-at-risk estimation errors.

Tags

# Anomaly Detection on Financial Time Series by Principal Component Analysis and Neural Networks


# Anomaly Detection on Financial Time Series by Principal Component Analysis and Neural Networks









A major concern when dealing with financial time series involving a wide variety ofmarket risk factors is the presence of anomalies. These induce a miscalibration of the models used toquantify and manage risk, resulting in potential erroneous risk measures. We propose an approachthat aims to improve anomaly detection in financial time series, overcoming most of the inherentdifficulties. Valuable features are extracted from the time series by compressing and reconstructingthe data through principal component analysis. We then define an anomaly score using a feedforwardneural network. A time series is considered to be contaminated when its anomaly score exceeds agiven cutoff value. This cutoff value is not a hand-set parameter but rather is calibrated as a neuralnetwork parameter throughout the minimization of a customized loss function. The efficiency of theproposed approach compared to several well-known anomaly detection algorithms is numericallydemonstrated on both synthetic and real data sets, with high and stable performance being achievedwith the PCA NN approach. We show that value-at-risk estimation errors are reduced when theproposed anomaly detection model is used with a basic imputation approach to correct the anomaly.

Shown in full with attribution under the source's licence. Licence: abstract CC0

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.