Random ForestIt is a multifunctional algorithm that includes multiple decision trees. It uses a sample set consisting of replacement sampling to train the decision tree. Each node of the decision tree only uses some features of random sampling during training.
When classifying a new object based on an attribute, each tree in the random forest will first give its own classification choice and "vote" for it. For classification problems, the output of the forest will be the one with the most votes; for regression problems, the output of the forest will be the average of the outputs of the decision trees.
In the random forest algorithm, "random" is the core, and "forest" is just a combination method. When constructing each tree, the forest usually adopts two to three layers of randomness to ensure the independence of each tree.
Random Forest Features
- Advantages: extremely high accuracy, not easy to overfit, good noise resistance, can handle high-dimensional data without feature selection, can handle discrete data and continuous data, data sets do not need to be normalized, fast training speed, can get variable importance ranking, and easy to parallelize.
- Disadvantages: The parameters are complex, a large amount of space and time are required for training, and some areas of the model cannot be explained.
Random Forest Application
- Perform regression and classification tasks;
- Used to handle missing values, outliers, and other important steps in data exploration;
- Used to combine several inefficient models into one efficient model.
Parent word: Bagging algorithm
Sub-word: Decision Tree
References
【1】https://blog.csdn.net/qq547276542/article/details/78304454
【2】https://blog.csdn.net/lishuandao/article/details/52555103
【3】https://en.wikipedia.org/wiki/Random_forest
【4】http://dataunion.org/23602.html