9. Save / Load M.L Model [파이썬으로 배우는 데이터 사이언스]

Data-Learning.tistory.com - 머신러닝 오픈소스 무료 강좌

---

 

 

Save/Load M.L Model

 

 

 

 

Contents: 

 

1. Pickle 

2. Joblib 

 

 

---

 

 

1. Pickle 

 

 

• pickle을 통하여 머신 러닝 알고리즘을 serialize 하고 파일에 저장할 수 있다. 이후 파일을 모델에 deserialize 하여 새로운 데이터에 사용할 수 있다. 

 

from sklearn.model_selection import train_test_split 

from sklearn.linear_model import LogisticRegression 

from pickle import dump, load 

X_train, X_test, y_train, y_test=train_test_split(X, y, test_size=0.3, random_state=6) 

 

---

 

#Fit model  

model=LogisticRegression(solver='liblinear') 

model.fit(X_train,y_train) 

 

---

 

#save model to disk 

filename='final_model.csv' 

dump(model, open(filename, 'wb')) 

 

 

---

#load model from disk 

loaded_model=load(open(filename, 'rb')) 

result=loaded_model.score(X_test, y_test) 

print(result) 

 

 

 

 

---

 

2. Joblib 

 

• joblib은 Scipy ecosystem중 하나이고 파이프라인을 제공한다. 
• 이것은 save/load에 관하여 numpy data structure을 사용하여 효과적으로 object을 만든다. 
• 많은 수의 parameters을 사용하거나 모든 데이터셋을 저장할 때, 효과적이다. 

from sklearn.model_selection import train_test_split 

from sklearn.linear_model import LogisticRegression 

from sklearn.externals.joblib import dump, load 

X_train, X_test, y_train, y_test=train_test_split(X, y, test_size=0.3, random_state=6) 

 

---

 

#Fit model  

model=LogisticRegression(solver='liblinear') 

model.fit(X_train,y_train) 

 

---

 

#save model to disk 

filename='final_model.csv' 

dump(model, filename) 

 

---

 

#load model from disk 

loaded_model=load(filename) 

result=loaded_model.score(X_test, y_test) 

print(result) 

 

 

 

---

밑바닥부터 시작하는 데이터 러닝

* 해당 게시물은 첫번째 배너 및 출처를 첨부하면 상업적 용도를 제외하고 사용하실수 있습니다.