Table 3.

Brief summary of the main characteristics of the different methods used in ML.

Method	Main Characteristics
Regression	Evaluation of the relationships between input variables and associated outputs and modeling of the relationship between them. Use of continous values. Linear regression: the simplest form, the basic idea is simply finding a line that best fits the data. Multiple linear regression and polynomial regression: focus on non-linear problems Logistic regression: models the probability of an observation to belong to a finite number of classes, typically two (0 and 1).
Classification	Finding of a model or function which helps in separating the data into classes based on different parameters. Use of discrete values. Categorization of data under different labels, according to some parameters given in input
Support Vector Machine (SVM)	Classification algorithm based on a hyperplane space that linearly separates training observations of different classes and creates a demarcation among the categories. Every unseen sample is classified into one of the classes, depending on the side on which it appears. Data that cannot be separated by a single continuous hyperplane are usually transformed using the kernel functions.
Decision Tree	Tree-like support tools used to correspond to a cause and its effect. Each node of the tree represents a test of one or more features of the observation and determines the following nodes to go through. The last nodes of the decision tree, where a decision is taken, are defined leaves of the tree The more nodes are present, the more accurate the decision tree will be. It can use regression or classification algorithms.
Random Forest	Combination of multiple decision trees, usually resulting in an improved predictive performance. Use of an “ensemble learning methods” (methods that use multiple learning algorithms to obtain better predictive performance than any of the constituent learning algorithm alone). Efficient modeling of complex and nonlinear data types, overcoming the limitations of Decision Trees. It can use regression or classification algorithms.
Neural Network (NN)	Similarity to the biological neural network, it is a collection of connected nodes called “artificial neurons”, which, like in the synapses in a real brain, can transmit information to other nodes or “neurons”. It is a network of mathematical equations. It works on input variables and, by going through a network of equations, transforms them in one or more output variables. Networks are built up of layers, each responsible for a linear transformation, followed by a nonlinear activation function. There are an input layer, one or more hidden layers, and an output layer Generally, more nodes and more layers allow the neural network to make much more complex calculations. It can use regression and classification algorithms, or combinations of them.
Deep Neural Networks (DNNs)	NNs with multiple hidden layers between the input and output layers.
Convolutional Neural Networks (CNN)	Its architecture is analogous to that of the connectivity pattern of neurons in visual cortex of the human brain. The hidden layers include layers that perform convolutions (in mathematics convolution is a mathematical operation on two functions that produces a third function that expresses how the shape of one is modified by the other).