https://towardsdatascience.com/5-ways-to-detect-outliers-that-every-data-scientist-should-know-python-code-70a54335a623
https://towardsdatascience.com/dbscan-algorithm-complete-guide-and-application-with-python-scikit-learn-d690cbae4c5d

Ways of detecting outliers

Isolation forests isolate outliers found using anomaly detection. This is adverse to other techniques that typically use data profiling (which are statistical methods used to describe the data set as a whole).


Isolation  Forest

The boxplot method is a statistical approach for detecting outliers that visualizes the distribution and spread of data. It identifies anomalies by marking any data points that fall beyond the whiskers of the plot.

Boxplot method for outlier detection

A robust random cut forest is an anomaly detection algorithm that isolates outliers through association. It assigns an anomaly score to data points, categorizing them as 'normal', 'low', or 'high'. The algorithm is effective for identifying anomalies in both offline and streaming data.

Robust random cut forest algorithm

- Isolation forest
- Robust random cut forest 
- Boxplots/Statistical Approach


University of Michigan - Ann Arbor

Finding unusual patterns in the data

https://www.sciencedirect.com/topics/computer-science/outlier-detection

Outlier detection

Outlier: data that differ dramatically from all others and distinguish themselves in characteristics.

Method 1 — Standard Deviation: data point that is more than 3 times the standard deviation, then those points are very likely to be outliers

Method 2 — Boxplots: Any data points that show above or below the whiskers, can be considered outliers.

Method 3— DBScan Clustering: Three core concepts (points) in DBScan:
Core Points
Border Points
Everything else is called Noise Points

(https://towardsdatascience.com/5-ways-to-detect-outliers-that-every-data-scientist-should-know-python-code-70a54335a623)

3 Methods to detect Outlier. 

WHAT IS OUTLIER DETECTION: identification of points that are extreme in value

Outlier: large residual / departure in the vertical dimension

Leverage point: the extreme value of predictor variable / big difference in the horizontal dimension

Influence point: Removing observation substantially by changing estimate of coefficient

Outliers, Leverage Points and Influential Points


Which of the following methods is most suitable for anomaly detection in n (n > 1) dimensional space？

Learn Before

Related

Learn After