[PYTHON] A memorandum of method often used in machine learning using scikit-learn (for beginners)

Introduction

We have summarized the methods that are often used when doing machine learning. We will make corrections as needed.

Preprocessing

Standardization

StandardScaler

from sklearn.preprocessing import StandardScaler

scaler = StandardScaler()  #Instance creation
scaler.fit(pd_sample)      #Parameter calculation (mean, standard deviation, etc.)
pd_sample_sc = scaler.transform(pd_sample)  #Data conversion

#pd_sample_sc = scaler.fit_transform(pd_sample)Can be executed collectively with

Dummy variable

get_dummies

#pandas.get_dummies()function
pd_sample = pd.get_dummies(pd_sample)

Training data / evaluation data division

train_test_split

from sklearn.model_selection import train_test_split

X_train, X_test, y_train, y_test = train_test_split(X, y)

Unsupervised learning

Clustering

KMeans

from skleran.cluster import KMeans

kmeans = KMeans(n_clusters=4, random_state=0)  #K-means model definition
clusters = kmeans.fit(pd_sample)               #Clustering execution
pd_sample['cluster'] = clusters.labels_        #Get clustering results

Dimensionality reduction

PCA

from sklearn.decomposition import PCA

pca = PCA(n_components=2)         #PCA model definition
pca.fit(pd_sample)                #Principal component analysis
x_pca = pca.transform(pd_sample)  #Data conversion (return value is array type object)
x_pca = pd.DataFrame(x_pca)       #Restore in DataFrame type

#x_pca = pca.fit_transform(pd_sample)Can be executed collectively with

Supervised learning

Regression model

LinearRegression

from sklearn.linear_model import LinearRegression()

model = LinearRegreession()  #Model initialization
model.fit(X_train, y_train)  #Modeling

#Accuracy verification of learning data and evaluation data
print(model.score(X_train, y_train))
print(model.score(X_test, y_test))

#Outputs a coefficient representing the degree of contribution for each explanatory variable
coef = pd.DataFrame({"feature_names":X.columns, "coefficient":model.coef_})
print(coef)

#Predict regression values for unknown data
print(model.predict(x_pred))

Classification model

DecisionTreeClassifier

from sklearn.tree import DecisionTreeClassifier

model = DecisionTreeClassifier(random_state=0)  #Model initialization
model.fit(X_train, y_train)                     #Modeling

#Accuracy verification of learning data and evaluation data
print(model.score(X_train, y_train))
print(model.score(X_test, y_test))

#Outputs a coefficient representing the degree of contribution for each explanatory variable
importance = pd.DataFrame({"feature_names":X.columns, "coefficient":model.feature_importances_})
print(importance)

#Predict classification values for unknown data
print(model.predict(x_pred))

#0/Output the prediction probability of 1
print(model.predict_proba(x_pred))

Validation of classification model

#Correct answer rate= (TP+TN)/(TP+FN+FP+TN)
model.score(X_test, y_test)

#Mixed matrix
from skleran.metrics import confusion_matrix
matrix = confusion_matrix(X_test, y_test)

#Heat map of mixed matrix
import seaborn as sns
sns.heatmap(matrix, annot=True, cmap='Blues')
plt.xlabel('Prediction')
plt.ylabel('Target')
plt.show()

#Adaptation rate= TP/(TP+FP)
from sklearn.metrics import precision_score
precision_score(X_test, y_test)

#Recall= TP/(TP+FN)
from sklearn.metrics import recall_score
recall_score(X_test, y_test)

#F value= 2*(Precision*Recall)/(Precision+Recall)
from sklearn.metrics import f1_score
f1_score(X_test, y_test)