Overview of classification algorithms in River
There is a large number of online classification models available in the River online machine learning package.
A selection of relevant ones is as follows:
LogisticRegressionPerceptronAdaptiveRandomForestClassifierALMAClassifierPAClassifier
Classification algorithm 1 – LogisticRegression
Logistic regression is one of the most basic statistical classification models. It models a dependent variable (target variable) that has two classes (1 or 0) and can use multiple independent variables to make the prediction.
The model combines each of the independent variables as log-odds; you can see this as the coefficients in linear regression, except that they are log-odds for each variable. The split in the model is based on the logistic function.
You can see a simplified schematic of the idea as follows:
Figure 6.4 – The logistic curve
