Supervised and Unsupervised machine learning

Supervised learning

Machine learning is focused on modelling, having several types for example supervised learning or unsupervised learning.

Supervised learning is divided on:

• Classification: Assigning categories to observation, predicting discreet variables that can have limited results.

For example: getting access to the university we can select students by GPA (<X accepted, >X rejected). Because we use only two features, we are able to plot and interpret the results quite easy without any machine learning, but if we get more variables we could use a support vector machine that will measure the different elements (for example High School GPA + Extracurricular + languages) and then filter people by using linear or polynomial classifiers.

• Regression: Continuous variables can take any value (worth, mass, height and others)

Unsupervised learning

Similar supervised learning but without a target column, not having a guide and looking directly at the whole dataset and trying to detect patterns. You can find insights without knowing too much about the dataset.

This methodology is used in 3 different cases:

• Clustering: Identifies groups in the dataset. it can compare it within its members of the group and as well similarities with other groups. Some clustering models, such as K Means need you to specify the nº of clusters, in others, DBSCAN (density-based clustering of applications with noise), requires the user to specify what constitutes a cluster.
• Anomaly detection: Detecting outliers or observations that differ from the others.
• Association: Finding relationships between elements.