MODULE 2 S3.2.2 Flashcards
Random Forest (Supplementary)
Decision trees can grow many branches until each split is as __________ as possible.
pure
____________________ are a popular model that mitigates this problem of overfitting in decision trees.
Random Forests
Alternatives to Using a Decision Tree
Random forest
Gradient boosting machine
Support vector machine
Neural network
It is a popular machine learning algorithm that merges the outputs of numerous decision trees to produce a single outcome.
It is also suitable for both classification and regression tasks.
Random Forest
Random forest was first introduced by ____________ and ____________ in ________
Leo Breiman
Adele Cutler
2001
The foundational ideas of Random forest date back to ________, when ___________ and _____________ proposed a method using randomized decision trees.
1993
Salzberg
Heath
The first algorithm for random decision forests was created by ____________ in __________ using the random subspace method.
Tin Kam Ho
1995
T/F Random Forest is suitable for classification tasks.
FALSE
Classification and Regression tasks
Random Forest’s strength lies in its ability to handle ________________ and mitigate ______________
complex dataset
overfitting
Random forest technique
Bagging
A key concept of random forest that combines the predictions of several base estimators to improve generalizability and robustness.
Ensemble Learning
A key concept of random forest which states that, at each split in the tree, a ransom subset of features is considered for splitting.
Random Feature Selection
Two stages of Random Forest:
Bootstrapping Stage
Splitting Stage
Considers all features to create different data samples (RF stage)
Bootstrapping Stage
Considers a random subset of features at each split (RF stage)
Splitting Stage
It is a method where multiple machine learning models are trained to solve the same problem and then combined to improve the final output.
Ensemble learning
It is a collection of models is used to make predictions rather than an individual model.
Ensemble learning
Two types of ensemble methods/techniques
Bagging
Boosting
Ensemble Process
Bagging : _______________
Boosting : _______________
Parallel
Sequential
Bagging is known as _____________
Bootstrap Aggregation
It serves as the ensemble technique in the Random Forest algorithm. It is a method of generating a new dataset with a replacement from an existing dataset.
Bagging
One of the famous techniques used in Bagging is _______________
Random Forest
How does bagging work?
- Sampling
- Independent training
- Aggregation
Prediction (Bagging Aggregation)
Classification : ______________
Regression : _______________
majority voting
average
Strengths of Bagging
Reduces Variance
Parallelism
Another ensemble technique, but unlike Bagging, it focuses on reducing bias by sequentially training models.
Boosting
T/F In boosting, the models created are independent.
False
T/F In boosting, the models are not independent; they are built one after another, with each model trying to fix the mistakes of the previous ones.
True
It refers to the family of an algorithm that converts weak learners (base learners) to strong learners.
Boosting
classifiers that are correct only up to a small extent with the actual classification
weak learner
Weak learner = _______________ = ______________
Base learner
Subtree
classifiers that are well correlated with the actual classification
strong learner
How does boosting work?
- Sequential training
- Weight adjustment
- Final model
Strengths of Boosting
Reduces Bias
Higher Accuracy
Two types of hyperparameters of random forest
Increase the Predictive Power
Increase the Speed
Predictive power parameters
n_estimators
max_features
mini_sample_leaf
criterion
max_leaf nodes
Speed parameters
n_jobs
random_state
oob_score
RF hyperparameter that determines the number of trees the algorithm builds before averaging the predictions
n_estimators
RF hyperparameter that pertains to the maximum number of features random forest considers splitting a node
max_features
RF hyperparameter that determines the minimum number of leaves required to split an internal node
mini_sample_leaf
RF hyperparameter:
How to split the node in each tree? (Entropy, Gini impurity, Log loss)
criterion
RF hyperparameter:
Maximum leaf nodes in each tree
max_leaf_nodes
RF hyperparameter:
It tel the engine how many processors it is allowed to use.
Processors:
1 : ______________
-1 : ______________
n_jobs
One processor
No limit
RF hyperparameter:
Controls randomness of the sample
random_state
RF hyperparameter:
It is a random forest cross-validation method
oob_score
(OOB : out of the bag)