SKlearn ‫ مكتبة‬:‫القسم العاشر‬

A. Data Preparation 10. Ensemble Classifier

1. Data files from SKlearn 11. K Nearest Neighbors
2. Data cleaning 12. Naïve Bayes
3. Metrics module 13. LDA , QDA
4. Feature selection 14. Hierarchical Clusters
5. Data Scaling 15. DbScan
6. Data Split 16. NLP
17. Apriori
B. ML Algorithms
1. Linear Regression C. Algorithm Evaluation :
2. Logistic Regression 1. Model Check
3. Neural Network 2. Grid Search
4. SVR 3. Pipeline
5. SVC 4. Model Save
6. K-means
7. PCA D. Time Series
8. Decision Tree
9. Ensemble Regression

1
 1.4 ) Feature Selection
, y ‫ المطلوبة و المؤثرة و استبعاد الباقين ويتم اختيارها بناء علي مدي ارتباطها بالمخرج‬features ‫و هي خاصة باختيار الـ‬
feature_selection ‫وتتم عبر الموديول‬

1.4.1 feature_selection.SelectPercentile
1.4.2 feature_selection.GenericUnivariateSelect
1.4.3 feature_selection.SelectKBest
1.4.4 feature_selection.SelectFromModel

2
 1.4.1) Select Percentile
, ‫ ويتم تحديد مقدار االهمية بطرق عديدة‬, ‫ واللي بتختار اهم فيتشرز مرتبطة بالناتج حسب النسبة المئوية المعطاة‬selectpercentile ‫يتم استخدام اداة‬
chi2 ‫ او أداة‬f_classif ‫مثل أداة‬

: ‫الصيغة‬
# Import Libraries
from sklearn.feature_selection import SelectPercentile
from sklearn.feature_selection import chi2 , f_classif
#----------------------------------------------------

#Feature Selection by Percentile

print('Original X Shape is ' , X.shape)
FeatureSelection = SelectPercentile(score_func = chi2, percentile=20) # score_func can = f_classif
X = FeatureSelection.fit_transform(X, y)

#showing X Dimension

3
print('X Shape is ' , X.shape)
print('Selected Features are : ' , FeatureSelection.get_support())

‫مثال‬
from sklearn.datasets import load_digits
from sklearn.feature_selection import SelectPercentile, chi2
X, y = load_digits(return_X_y=True)
64 ‫هنا يتم اظهار عدد فيتشرز الداتا و هي‬
X.shape

‫ فيشترز‬7 ‫ منهم اي‬% 10 ‫نقوم بجعله يختار اهم‬

X_new = SelectPercentile(score_func =chi2, percentile=10).fit_transform(X, y)

print(X_new.shape)

‫مثال آخر‬

from sklearn.datasets import load_breast_cancer

from sklearn.feature_selection import SelectPercentile , chi2

4
data = load_breast_cancer()
X = data.data
y = data.target
X.shape
sel = SelectPercentile(score_func = chi2 , percentile = 20).fit_transform(X,y)
sel.shape

‫و اذا اردنا معرفة الفيتشرز المختارة و المستبعدة‬

from sklearn.datasets import load_digits
from sklearn.feature_selection import SelectPercentile, chi2
X, y = load_digits(return_X_y=True)
X.shape

X_new = SelectPercentile(score_func =chi2, percentile=10)

X_new.fit(X, y)
selected = X_new.transform(X)
X_new.get_support()

5
 1.4.2) Generic Univariate Select
‫و فيها يتم اختيار عدد معين من الفيتشرز بناء علي احد االدوات‬

: ‫الصيغة‬
#Import Libraries
from sklearn.feature_selection import GenericUnivariateSelect
from sklearn.feature_selection import chi2 , f_classif
#----------------------------------------------------
#Feature Selection by Generic
#print('Original X Shape is ' , X.shape)
FeatureSelection = GenericUnivariateSelect(score_func= chi2, mode= 'k_best', param=3) # score_func can =
f_classif : mode can = percentile,fpr,fdr,fwe
X = FeatureSelection.fit_transform(X, y)
#showing X Dimension
#print('X Shape is ' , X.shape)
#print('Selected Features are : ' , FeatureSelection.get_support())

6
 ‫ فيتشرز‬5 ‫وهنا يتم تحديد عدد‬

from sklearn.datasets import load_breast_cancer

from sklearn.feature_selection import GenericUnivariateSelect, chi2
X, y = load_breast_cancer(return_X_y=True)
X.shape

transformer = GenericUnivariateSelect(chi2, 'k_best', param=5)

X_new = transformer.fit_transform(X, y)

X_new.shape

transformer.get_support()

7
 1.4.3) Select KBest
‫ باسلوب رياضي مختلف‬features ‫يقوم كذلك باختيار عدد معين من الـ‬

: ‫الصيغة‬
#Import Libraries
from sklearn.feature_selection import SelectKBest
from sklearn.feature_selection import chi2 , f_classif
#----------------------------------------------------
#Feature Selection by KBest
#print('Original X Shape is ' , X.shape)
FeatureSelection = SelectKBest(score_func= chi2 ,k=3) # score_func can = f_classif
X = FeatureSelection.fit_transform(X, y)

#showing X Dimension
#print('X Shape is ' , X.shape)
#print('Selected Features are : ' , FeatureSelection.get_support())

8
 ‫مثال‬
from sklearn.datasets import load_digits
from sklearn.feature_selection import SelectKBest, chi2
X, y = load_digits(return_X_y=True)
X.shape

X_new = SelectKBest(chi2, k=30).fit_transform(X, y)

X_new.shape

9
 1.4.4) Select From Model
selectfrommodel ‫ وده بامر‬, ‫يتم اختيار الفيتشر بناء علي موديل معين بحيث الموديل نفسه يشوف انهي فيشترز مهمة‬

: ‫الصيغة‬
#Import Libraries
from sklearn.feature_selection import SelectFromModel
#----------------------------------------------------

#Feature Selection by KBest

#print('Original X Shape is ' , X.shape)

'''
from sklearn.linear_model import LinearRegression
thismodel = LinearRegression()
'''

10
FeatureSelection = SelectFromModel(estimator = thismodel, max_features = None) # make sure that thismodel is
well-defined
X = FeatureSelection.fit_transform(X, y)

#showing X Dimension
#print('X Shape is ' , X.shape)
#print('Selected Features are : ' , FeatureSelection.get_support())

‫مثال‬
from sklearn.datasets import load_breast_cancer
from sklearn.ensemble import RandomForestClassifier
from sklearn.feature_selection import SelectFromModel
data = load_breast_cancer()
X = data.data
y = data.target
‫ وعلي اساسها هيتم اختيار اقوي فيتشرز‬, ‫ لها‬20 ‫ برقم‬, ‫تم تحديد موديل الغابة العشوائية‬

sel = SelectFromModel(RandomForestClassifier(n_estimators = 20))

11
‫)‪sel.fit(X,y‬‬
‫)‪selected_features = sel.transform(X‬‬
‫)(‪sel.get_support‬‬

‫الحظ ان مش الزم يتم اختيار نفس الموديل في الترين ‪ ,‬ممكن موديل تاني ‪ ,‬فاختيار الفيتشرز عادي من موديل مختلف ‪ ,‬كمان متنساش ان ممكن يتم‬
‫عمل خطوات ورا بعص ‪ ,‬يعني مثال بولينوميال عشان اعمل فيتشرز كتير جدا ‪ ,‬بعدها اجيب موديل يختار منهم‬

‫‪12‬‬