機械学習・データサイエンスのアウトプット

# 使用するライブラリをインポート
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt

import io
from google.colab import files

from sklearn.model_selection import train_test_split
from keras.models import Sequential
from keras.layers import Dense, Activation
from keras.optimizers import Adam
from keras.layers.core import Dropout
from keras.callbacks import EarlyStopping, LearningRateScheduler

from sklearn.metrics import mean_absolute_error
from sklearn.metrics import mean_squared_error
from math import sqrt
from sklearn.metrics import r2_score


# Google Colaboratoryでファイルをアップロード
uploaded = files.upload()


# データフレームとして、変数dfに3列×10000行のデータを格納
df = pd.read_csv(io.BytesIO(uploaded['data10000.csv']))


# データの分割
(train, test) = train_test_split(df, test_size=0.2, shuffle=True)

# x_train, x_testではcsvファイルの1列目から2列目を，y_train, y_testでは3列目のデータを活用
x_train = train.iloc[:, [0, 1]]
y_train = train.iloc[:, [2]]

x_test = test.iloc[:, [0, 1]]
y_test = test.iloc[:, [2]]


# データの正規化
x_train = (x_train - x_train.min()) / (x_train.max() - x_train.min())
x_test = (x_test - x_test.min()) / (x_test.max() - x_test.min())


# モデルの構築
# inputの数
n_in = 2

# ノードの数
n_hidden = 16

# outputの数
n_out = 1

# 学習回数
epochs = 50
batch_size = 64

model = Sequential()
model.add(Dense(n_hidden, activation= 'relu', input_dim=n_in))
model.add(Dense(n_hidden, activation= 'relu', input_dim=n_in))
model.add(Dense(n_hidden, activation= 'relu', input_dim=n_in))
model.add(Dense(units=n_out))
model.summary()

optimizer = Adam(lr=0.001, beta_1=0.9, beta_2=0.999)
model.compile(loss='mean_squared_error', optimizer=optimizer)


# 学習オプション
history = model.fit(x_train, y_train, epochs=epochs, batch_size=batch_size, validation_data=(x_test, y_test))


# 構築したモデルで予測
test_predict = model.predict(x_test)


#精度評価指標
# MAE
test_MAE = mean_absolute_error(y_test, test_predict)
print('MAE：', test_MAE)

# MSE
test_MSE = mean_squared_error(y_test, test_predict)
print('MSE：', test_MSE)

# RMSE
test_RMSE = sqrt(mean_squared_error(y_test, test_predict))
print('RMSE：', test_RMSE)

# 決定係数
R2 = r2_score(y_test, test_predict)
print('R2：', R2)

# MAEに対するRMSEの比
print('RMSE/MAE：', test_RMSE / test_MAE)


# 損失関数
loss = history.history['loss']
val_loss = history.history['val_loss']
epochs = len(loss)
plt.plot(range(epochs), loss, marker='.', label='loss')
plt.plot(range(epochs), val_loss, marker='.', label='val_loss')
plt.legend(loc='best')
plt.grid()
plt.xlabel('epoch')
plt.ylabel('loss')
plt.show()


# Observed-Predicted Plot
plt.figure()
plt.scatter(y_test, test_predict, c='blue', alpha=0.8)
plt.ylim(plt.ylim())
plt.grid()
plt.xlabel('observed_data')
plt.ylabel('predict_data')
plt.show()

MAE： 0.02991514426947141   # 0に近いほど良い
MSE： 0.001563834395335399   # 0に近いほど良い
RMSE： 0.03954534606417548   # 0に近いほど良い
R2： 0.9988363429433301   # 1に近いほど良い
RMSE / MAE： 1.3219172773481072   # 1.253に近いほど良い