Random Forest Tree

A version of Ensemble Learning.

Ensemble Learning: Take multiple algorithms and use them together.

Steps Involved:

1. Pick at random K data points from the Training set.
   • Using the entire dataset, we pick k data points from that set
2. Build the Decision Tree associated to these K data points.
   • Rather than building a decision tree based on everything in your dataset. We just build it based of those K data points.
3. Choose the number of Ntree of trees you want to build and repeat Step 1 and step 2.
4. For a new data point, make each one of your Ntree trees predict the value of Y to the data point in question, and assign the new data point the average across all the predicted Y values.
   • Basically, instead of getting one prediction from a Decision Tree we get many (usually 500 trees) and we take the average across those.

Why we use it

The reason for using Ensemble Learning is because one algorithm is typically not too accurate or may have major flaws in some respects. By using many algorithms and averaging their results we reduce errors. Because changes in the dataset can greatly affect one algorithm that could ruin the prediction but with many algorithms being used it is much harder for results to get skewed, making it very durable.

Building a Random Forest Regression

Use the regression template to set everything up. Use Higher resolution visualization.

Create Regressor

Import the RandomForestRegressor class from the library sklearn.ensemble n_estimators: Number of trees in forest

from sklearn.ensemble import RandomForestRegressor
regrssor = RandomForestRegressor(n_estimators = 300, random_state = 0)
regressor.fit(X, y)