top of page
Search

Outstanding Outliers: Why You Should Care About Anomalies


Why Are We Interested in Anomalies?


Anomaly detection is the process of identifying unexpected items or events in data sets, which differ from the norm. Anomaly detection identifies data points, events, and/or observations that deviate from a dataset’s normal behaviour. Anomalous data can indicate critical incidents, such as a technical glitch or a major financial event. For example, the 1929 Wall Street crash was an anomalous event.



Now you must be wondering what is the point of detecting outliers/anomalies? After all, the impacts of the Wall Street crisis were felt across the world. Global-scale 'anomalies' such as the Wall Street crisis of 1929 or the subprime crisis of 2008 are not the only anomalies that exist in the financial world. Indeed recurring anomalies especially those of cyclical or a time-specific nature can have an impact on trading and investment tragedies. For instance, anomalies such as the 'weekend effect', 'January effect' have a noticeable effect on stock prices. News-related anomalies such as news of mergers and the stock split effect get reflected in the stock prices. Identifying these anomalies and the anomalies arising from malpractices such as stock price manipulation can help inform trading decisions. Potential malpractices such as pump-and-dump, spoofing too can be detected via anomaly detection making this an important tool for those interested in financial markets.


What is Anomaly Detection?


In machine learning, anomaly detection is often applied on unlabeled data which is known as unsupervised anomaly detection. Autoencoders, commonly used for anomaly detection are a neural network approach to learning the specifics of the underlying dataset in an unsupervised manner Autoencoder models then learn the patterns of the input data to identify anomalies. Here is a jargon-free and fuss-free introduction to what autoencoders are:





 
 
 

Comments


Check back soon
Once posts are published, you’ll see them here.
FEATURED POSTS
RECENT POSTS

JOIN OUR MAILING LIST!

Thanks for submitting!

  • LinkedIn
  • Facebook
  • Medium_circle-512
  • circle-cropped
  • 25231
  • Twitter
  • YouTube

Copyright 2021 DataOwl | All Rights Reserved

bottom of page