# 확률적 경사 하강법¶

## SGDClassifier¶

In [1]:
import pandas as pd


In [2]:
fish_input = fish[['Weight','Length','Diagonal','Height','Width']].to_numpy()
fish_target = fish['Species'].to_numpy()

In [3]:
from sklearn.model_selection import train_test_split

train_input, test_input, train_target, test_target = train_test_split(
fish_input, fish_target, random_state=42)

In [4]:
from sklearn.preprocessing import StandardScaler

ss = StandardScaler()
ss.fit(train_input)
train_scaled = ss.transform(train_input)
test_scaled = ss.transform(test_input)

In [5]:
from sklearn.linear_model import SGDClassifier

In [6]:
sc = SGDClassifier(loss='log', max_iter=10, random_state=42)
sc.fit(train_scaled, train_target)

print(sc.score(train_scaled, train_target))
print(sc.score(test_scaled, test_target))

0.773109243697479
0.775

/usr/local/lib/python3.6/dist-packages/sklearn/linear_model/_stochastic_gradient.py:557: ConvergenceWarning: Maximum number of iteration reached before convergence. Consider increasing max_iter to improve the fit.
ConvergenceWarning)

In [7]:
sc.partial_fit(train_scaled, train_target)

print(sc.score(train_scaled, train_target))
print(sc.score(test_scaled, test_target))

0.8151260504201681
0.825


## 에포크와 과대/과소적합¶

In [8]:
import numpy as np

sc = SGDClassifier(loss='log', random_state=42)

train_score = []
test_score = []

classes = np.unique(train_target)

In [9]:
for _ in range(0, 300):
sc.partial_fit(train_scaled, train_target, classes=classes)

train_score.append(sc.score(train_scaled, train_target))
test_score.append(sc.score(test_scaled, test_target))

In [10]:
import matplotlib.pyplot as plt

plt.plot(train_score)
plt.plot(test_score)
plt.xlabel('epoch')
plt.ylabel('accuracy')
plt.show()

In [11]:
sc = SGDClassifier(loss='log', max_iter=100, tol=None, random_state=42)
sc.fit(train_scaled, train_target)

print(sc.score(train_scaled, train_target))
print(sc.score(test_scaled, test_target))

0.957983193277311
0.925

In [12]:
sc = SGDClassifier(loss='hinge', max_iter=100, tol=None, random_state=42)
sc.fit(train_scaled, train_target)

print(sc.score(train_scaled, train_target))
print(sc.score(test_scaled, test_target))

0.9495798319327731
0.925