Random forests is a supervised learning algorithm. It can be used both for classification and regression. It is also the most flexible and easy to use algorithm. A forest is comprised of trees. It is said that the more trees it has, the more robust a forest is. Random forests creates decision trees on randomly selected data samples, gets prediction from each tree and selects the best solution by means of voting. It also provides a pretty good indicator of the feature importance.

Random forests has a variety of applications, such as recommendation engines, image classification and feature selection. It can be used to classify loyal loan applicants, identify fraudulent activity and predict diseases. It lies at the base of the Boruta algorithm, which selects important features in a dataset.

The individual decision trees are generated using an attribute selection indicator such as information gain, gain ratio, and Gini index of impurity for each attribute. Each tree depends on an independent random sample. In a classification problem, each tree votes and the most popular class is chosen as the final result. In the case of regression, the average of all the tree outputs is considered as the final result. It is simpler and more powerful compared to the other non-linear classification algorithms.

In this case, we are going to apply Random Forest algorithm to the Breast Cancer dataset available in the sklearn datasets library to classify if it benign or malign tumor.

pip install RandomForest

git clone https://github.com/stkobsar/RandomForest.git
cd RandomForest
python setup.py install

python RandomForest/main.py 

• Confusion matrix

In order to evaluate how good is the model, we have to see if classifies correctly malign and benign tumors. We've got:

• 51 true positives

• 2 false positives

• 86 true negatives

• 4 false negatives

• Cross validation

A typical 10-fold cross validation is applied getting 95.31% accuracy

• Model accuracy

The model has 95.8% of accuracy in classification

Random Forest is a good algorithm to solve this problem, since presents quite similar accuracy each from prediction and validation, and has a very good rate of true positives and true negatives.