Statistical analysis and anomaly detection

Basic machine learning and statistical analysis

In this case, this is referring mostly to unsupervised learning, leveraged to identify patterns or maliciousness, without the need for training or intervention directly from the user. This includes approaches like anomaly or outlier detection. These can be configured as Elastic ML jobs.

Anomaly detection is built on time series data and uses a probability model, where the model evolves over time, which also allows for future forecasting.

Outlier detection does not require time series, which uses data frame analytics to perform clustering and relative distancing, helping identify outliers.

Advanced machine learning and statistical analysis

This encompasses mostly supervised learning, with approaches such as classification or regression.

Fig. 88 Machine learning workflow

Classification works by predicting distinct categorical values by understanding the relationships across the data.

Regression is also built on an understanding of relationships within the data, but focuses on predicting numerical values.