Quick manual sklearn. Decomposition. PCA parameters
call
Sklearn. Decomposition. PCA (ncomponents = None, copy = True, whiten = False, svdsolver = ‘auto’, tol = 0.0, iteratedpower = ‘auto’, N_components randomstate = None) parameters
paraphrase
The number of principal components n to be retained in PCA algorithm is also the number of retained features N
Set up the
Int or string, which defaults to None, all elements are retained. Assigning an int, such as =1, will reduce the raw data to a dimension of string, such as ‘mle’, and will automatically pick the number of features N so that the desired percentage of variance is satisfied
whiten
paraphrase
Whiten, so that each feature has the same variance.
Set up the
Bool. The default value is False. If there are subsequent data processing actions after PCA dimension reduction, whitening can be considered
svd_solver
paraphrase
Method of constant singular value decomposition SVD
Set up the
The Auto PCA class automatically selects the following three algorithm trade-offs
randomized
It is suitable for dimension reduction of PCA with large amount of data, multiple data dimensions and low proportion of principal components
full
SVD in the traditional sense uses the sciPY library counterpart
arpack
The SCIpy library’s SPARSE SVD was implemented directly, which was similar to randomized
copy
paraphrase
Indicates whether a copy of the original training data is made while running the algorithm.
Set up the
If True, the value of the original training data will not change after the PCA algorithm is run, because the operation is performed on the copy of the original data. If it is False, the value of the original training data will be changed after PCA algorithm is run, because dimension reduction is performed on the original data.
tol
Criteria for stopping solvers, type float, default value 0 When svd_solver selects’ arpack ‘, its error tolerance for running the SVD algorithm
iterated_power
Int or STR. The default value is’ auto ‘. Number of iterations of the SVD algorithm run when svd_solver selects’ randomized ‘
random_state
Int. Defaults to None. Seed of a pseudorandom number generator used for probability estimation when shuffling data
attribute
components_
Returns the component with the largest variance
explainedvariance
Variance of each principal component after dimensionality reduction. The larger the variance, the more important the principal component
explainedvarianceratio_
The proportion of variance value of each principal component to the total variance value after dimensionality reduction, the larger the proportion, the more important the principal component
singularvalues
Singular values of each of the corresponding components
mean_
Empirical mean of features estimated by training set = X.bean (Axis = 0)
ncomponents
Returns the number of retained ingredients n
nfeatures
Number of features of training data
nsamples
Sample number of training data
noisevariance
Returns the covariance of the noise
methods
fit(self, X[, y])
The PCA model was trained with data X
fit_transform(self, X[, y])
The PCA model is trained with X, and the data after dimension reduction is returned
get_covariance(self)
Calculate data covariance (with generative model)
get_params(self[, deep])
Get the parameters of PCA
get_precision(self)
Calculate data accuracy matrix (with generation model)
inverse_transform(self, X)
Convert the dimensionally reduced data to the original data, but it may not be exactly the same
score(self, X[, y])
Calculate the log likelihood average for all samples
score_samples(self, X)
Returns the logarithmic likelihood value for each sample
set_params(self, **params)
Set the parameters of PCA
transform(self, X)
Transform the data X into the data after dimension reduction. After the model is trained, the transform method can also be used to reduce the dimension of the newly input data
Machine Learning and Statistics [Welcome to scan code for attention and receive massive data][1]
[1]: https://img2018.cnblogs.com/blog/743008/201907/743008-20190725101305454-1314391880.png