File:Classifier.png

http://scikit-learn.org/stable/auto_examples/classification/plot_classifier_comparison.html

from the sklearn documentation. This is a comparison of different classifiers. trained on an arbitrary dataset showing the decision boundaries.

Here is a brief high-level description of each model.

nearest neighbors.Finds the points that are most dense and finds which output is most likely and uses those groups to classify.

Linear SVM. This draws vectors that are linear across the graph, cutting it into pieces. each piece is a different group. The kernel of the vector can be changed so as to get curved decision boundaries.

Decision trees: a tree is made that splits the data into two sets based on a query( is x <=30) and classifies each group differently. this is repeated with each variable until everything is classified.

Random Forest: Many, many trees are created that each classify the data in a different way. The confidence in each tree is asesed and the models are weighted and used together as an ensemble for the final model. These are very powerful models because they are stronger with more trees and there are many tricks to make them better

Adaboost: A random forest with a large amount of trees. it is sensitive to noise because it uses weak learners that dont have high probabilities to avoid over fitting.

Naive bayes: a Bayesian probability model is used. Each variable is considered to occur given each other variable occurs.each point is assigned a probability of occurring. the group with the highest probability is the classification.

Other models:

neural networks: a fully connected network is created where the data goes through multiple layers and is transformed. This can learn patterns that are hidden. This is a very powerful model with a few modifications. Each layer can have many nodes which creates the ability to learn complex patterns.