Types of Machine Learning Models

What is a model?

A model is a simplified representation of reality. The researcher always tries to develop a model that can either represent reality accurately or, at the very least, determine a set of conditions or assumptions that will allow them to control the testing of the model as scientists do in laboratories. 

In the context of machine learning, a model is the output of a machine learning algorithm run on data that represents what was learned by the machine learning algorithm.

What is the difference between supervised and unsupervised models? 

Supervised learning models infer patterns between a set of inputs, such as predictors or features, and the desired output response or target variable. These models are provided with historical data for both input and output variables in order to find the relationship that has the best predictive power for the sample data. One example would be a supervised linear logistic classification model that can predict the response “regime of the stock market” as a binary outcome “bull or bear” using as features a group of macroeconomic variables.

Unsupervised learning algorithms examine the dataset and identify relationships between variables and their common drivers. In other words, they try to unveil the structure of the data. As an example, we may take market returns and try to identify the main drivers of the market. A useful unsupervised model could classify stocks in groups depending on how close their embedded predictors move together, instead of using more rigid methodologies like region or industry classification.