Feature Selection
Feature SelectionIt is the process of selecting feature subsets, which is usually used to build models together. Its advantages are as follows:
- Simplify the model;
- Shorten training time;
- Improve generality and reduce overfitting
Feature selection algorithms can be viewed as a combination of search techniques and evaluation indicators. The former provides candidate new feature subsets, while the latter scores different feature subsets. The simplest one is to test each subset and find the subset with the lowest error rate. This algorithm requires an exhaustive search space. Since it is difficult to complete the calculation of all feature sets, it can only cover a small number of feature subsets, and has a greater impact on the algorithm.
Through different evaluation indicators, feature selection algorithms can be divided into three categories: packaging, filtering and embedding