Sklearn Feature Engineering (summary)

1、 Feature Engineering

     Dictionary feature extraction

from sklearn.feature_extraction import DictVectorizer  # Feature extracted packages     

 

   Text feature extraction and jieba participle

Text feature extraction , For example, document classification 、 Spam classification and news classification . Text classification is based on whether words exist 、 And the probability of words （ Importance ） To express .

If you want to count the number of Chinese words , It is necessary to segment Chinese words first .jieba

    tf-idf Text extraction

It is a commonly used weighting technique for information retrieval and text mining , This statistical method , Used to evaluate the importance of a word in a document .

from sklearn.feature_extraction.text import TfidfVectorizer    

Feature Engineering ~ normalization

normalization

X=（x-min)/(max-min)

among ,max and min Are the maximum and minimum values of a column respectively ,x Is the value before normalization .

from sklearn.preprocessing import MinMaxScaler

from sklearn.preprocessing import MinMaxScaler

scaler =MinMaxScaler()

data =[

    [180,75,35],[175,80,17],[159,50,46],[149,79,45]  

]

result =scaler.fit_transform(data)

print(result)

Standardization

from sklearn.preprocessing import StandardScaler # Standardization

scaler=StandardScaler()

result=scaler.fit_transform(data)

print(result)

Feature Engineering - Data dimension reduction

Principal component analysis

Principal component analysis , It is a statistical method . Through orthogonal transformation, a group of variables that may have correlation variables are transformed into a group of linearly uncorrelated variables , The transformed set of variables is called principal component .

The principal components need to remember two things ：

• The covariance between the features after dimensionality reduction is 0, Indicates that each feature relationship is independent , Each feature will not change regularly with the change of other features .

• The variance of each feature should be as large as possible .

from sklearn.decomposition import PCA

def pca_decomposition():

    pca=PCA(n_components=2)#1、0~1 Between , Scale of dimensions -1;2、 plastic ： Specific dimensions , It has to be for min(n_samples,n_features) within

    result =pca.fit_transform(

    [

        [4,2,76,9],

        [1,192,1,56],

        [34,5,20,90]

    ])

    print(result)

pca_decomposition()