Page 87 - Data Science Algorithms in a Week

4

Random Forest

A random forest is a set of random decision trees (similar to the ones described in the
previous chapter), each generated on a random subset of the data. A random forest
classifies the feature to belong to the class that is voted for by the majority of the random
decision trees. A random forest tends to provide a more accurate classification of a feature
than a decision tree because of the decreased bias and variance.

In this chapter, you will learn:

Tree bagging (or bootstrap aggregation) technique as part of random forest
construction, but that can be extended also to other algorithms and methods in
data science to reduce the bias and variance and hence to improve the accuracy
In example Swim preference to construct a random forest and classify a data item
using the constructed random forest
How to implement an algorithm in Python that would construct a random forest
In example Playing chess the differences in the analysis of a problem by
algorithms naive Bayes, decision trees and random forest
In example Going shopping how random forest algorithm can overcome the
shortcomings of decision tree algorithm and thus outperform it
In example Going shopping how a random forest can express the level of the
confidence in its classification of the feature
In exercises how decreasing the variance of a classifier can yield more accurate
results