Reduced dimension Reference URL http://dataunion.org/20803.html
"Low Variance filter" requires normalization of the data first
"High correlation filtering" thinks that when two columns of data change in a similar trend, they contain similar information
Random forest produces many large trees for the target attribute, and then finds the most informative subset of features based on the statistical results of each attribute. If an attribute is often the best split attribute, it is most likely the information feature that needs to be preserved
Principal component Analysis (PCA) requires normalization of the data first, and the core orthogonal transformation. The interpretation of data is lost after PCA transformation
"Reverse feature elimination (backward Feature elimination)"
"Forward feature construction (Forward Feature construction)"
Principal Component Analysis (PCA) reference URL http://blog.csdn.net/u012162613/article/details/42192293
A linear dimensionality reduction method that uses singular values to decompose and retain most useful information. Complete the singular value decomposition with scipy.linalg (only for square matrices and small data). Complexity of Time n^3
Parameters:
n_components(int, None or string)
Default N_components = = min(n_samples, n_features)
N_components = = ' Mle ' guess
copy : bool
False, the data passed into the training model will be overwritten with fit_transform (x) rather than fit (x). Transform (x)
whiten : bool, optional
False by default
True
Sklearn Study Notes