Ensemble algorithm
Ensemble learning (ensemble learning) is currently a very popular machine learning strategy, and basically all problems can be improved by borrowing its ideas. The basic starting point is to gather algorithms and various strategies together. To put it bluntly, it is a problem that everyone can join together! Integrated learning can be used for both classification and regression problems . It is often seen in the field of machine learning. This chapter discusses several classic integration strategies and interprets them in combination with their applications.
Bagging algorithm
The integration algorithm has three core ideas: bagging
, boosting
and stacking
, these integration strategies are still very easy to understand, and I will introduce them one by one below.
parallel integration
Bagging
That is boostrap aggregating
, where boostrap is a sampling method with replacement, and the sampling strategy is simple random sampling. The principle is very straightforward. Put multiple basic models together, and finally calculate the average value . Here, the decision tree can be used as the basic model. In fact, basically all integration strategies are based on the tree model . The formula is as follows:
f ( x