Anomalous activities can be linked to some kind of problems or rare events such as bank fraud, medical problems, structural defects, malfunctioning equipment etc.Įxample of an Anomalous Activity The Need for Anomaly DetectionĪccording to a research by Domo published in June 2018, over 2.5 quintillion bytes of data were created every single day, and it was estimated that by 2020, close to 1.7MB of data would be created every second for every person on earth. Anomaly detection (or outlier detection) is the identification of rare items, events or observations which raise suspicions by differing significantly from the majority of the data. Anomaly DetectionĪnomaly is a synonym for the word ‘outlier’. Here though, we’ll discuss how unsupervised learning is used to solve this problem and also understand why anomaly detection using unsupervised learning is beneficial in most cases. In each post so far, we discussed either a supervised learning algorithm or an unsupervised learning algorithm but in this post, we’ll be discussing Anomaly Detection algorithms, which can be solved using both, supervised and unsupervised learning methods. To consolidate our concepts, we also visualized the results of PCA on the MNIST digit dataset on Kaggle.
In the previous post, we had an in-depth look at Principal Component Analysis (PCA) and the problem it tries to solve. Detection of Fraudulent Transactions on a Credit Card Dataset available on Kaggle.Extending Baseline Algorithm for a Multivariate Gaussian Distribution and the use of Mahalanobis Distance.Evaluating an Anomaly Detection Algorithm.Baseline Algorithm for Anomaly Detection with underlying Mathematics.Unsupervised Learning For Anomaly Detection Contents
0 kommentar(er)
