Data Every Day: ライム病の発疹

tldr

KggleのLyme Disease RashesをLyme Disease Image Classification - Data Every Day #050に沿ってやっていきます。

実行環境はGoogle Colaboratorです。

インポート

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns

import sklearn.preprocessing as sp
from sklearn.model_selection import train_test_split
import sklearn.linear_model as slm
import sklearn.svm as svm
import sklearn.neural_network as snn

import tensorflow as tf

データのダウンロード

Google Driveをマウントします。

from google.colab import drive
drive.mount('/content/drive')

Mounted at /content/drive

KaggleのAPIクライアントを初期化し、認証します。
認証情報はGoogle Drive内（/content/drive/My Drive/Colab Notebooks/Kaggle）にkaggle.jsonとして置いてあります。

import os
kaggle_path = "/content/drive/My Drive/Colab Notebooks/Kaggle"
os.environ['KAGGLE_CONFIG_DIR'] = kaggle_path

from kaggle.api.kaggle_api_extended import KaggleApi
api = KaggleApi()
api.authenticate() 

Kaggle APIを使ってデータをダウンロードします。

dataset_id = 'sshikamaru/lyme-disease-rashes'
dataset = api.dataset_list_files(dataset_id)
file_name = dataset.files[0].name
file_path = os.path.join(api.get_default_download_dir(), file_name)
file_path

'/content/RashData/Train/Train_2_Cases/Lyme_Negative/Pityriasis rosea rash125.jpg'

api.dataset_download_files(dataset_id, force=True, quiet=False)

  0%|          | 0.00/84.5M [00:00<?, ?B/s]

Downloading lyme-disease-rashes.zip to /content


100%|██████████| 84.5M/84.5M [00:01<00:00, 67.3MB/s]

import zipfile

zip_path = os.path.join(api.get_default_download_dir(), '{}.zip'.format('lyme-disease-rashes'))
with zipfile.ZipFile(zip_path) as existing_zip:
    existing_zip.extractall(api.get_default_download_dir())

import os
os.listdir('/content/RashData')

['Lyme_Positive_By_Diease', 'Lyme_Negative_By_Disease', 'Train', 'Validation']

train_dir = '/content/RashData/Train/Train_2_Cases'
val_dir = '/content/RashData/Validation/Validation_2_Cases'

batch_size = 32
img_height = 128
img_width = 128

train_ds = tf.keras.preprocessing.image_dataset_from_directory(
    train_dir,
    validation_split=0.2,
    seed=42,
    subset='training',
    batch_size=batch_size,
    image_size=(img_height, img_width),
)

val_ds = tf.keras.preprocessing.image_dataset_from_directory(
    train_dir,
    validation_split=0.2,
    seed=42,
    subset='validation',
    batch_size=batch_size,
    image_size=(img_height, img_width),
)


test_ds = tf.keras.preprocessing.image_dataset_from_directory(
    val_dir,
    batch_size=batch_size,
    image_size=(img_height, img_width),
)

Found 357 files belonging to 2 classes.
Using 286 files for training.
Found 357 files belonging to 2 classes.
Using 71 files for validation.
Found 87 files belonging to 2 classes.

class_names = train_ds.class_names
print(class_names)

['Lyme_Negative', 'Lyme_Positive']

for images, labels in train_ds.take(1):
    print(images.shape)
    print(class_names[labels[0]])

(32, 128, 128, 3)
Lyme_Negative

バッチサイズ: 32
画像サイズ: 128*128(RGB)

可視化

plt.figure(figsize=(10, 10))
for images, labels in train_ds.take(1):
    for i in range(9):
        plt.subplot(3, 3, i + 1)
        plt.imshow(images[i].numpy().astype('uint8'))
        plt.title(class_names[labels[i]])
    plt.show()

モデリング

AUTOTUNE = tf.data.experimental.AUTOTUNE

train_ds = train_ds.cache().shuffle(1000).prefetch(buffer_size=AUTOTUNE)
val_ds = val_ds.cache().shuffle(1000).prefetch(buffer_size=AUTOTUNE)
test_ds = test_ds.cache().shuffle(1000).prefetch(buffer_size=AUTOTUNE)

model = tf.keras.Sequential([
    tf.keras.layers.experimental.preprocessing.Rescaling(1./255, input_shape=(img_height, img_width, 3)),
    tf.keras.layers.Conv2D(16, 3, padding='same', activation='relu'),
    tf.keras.layers.MaxPool2D(),
    tf.keras.layers.BatchNormalization(),
    tf.keras.layers.Conv2D(32, 3, padding='same', activation='relu'),
    tf.keras.layers.MaxPool2D(),
    tf.keras.layers.BatchNormalization(),
    tf.keras.layers.Flatten(),
    tf.keras.layers.Dense(128, activation='relu'),
    tf.keras.layers.Dense(1, activation='sigmoid'),
])

model.summary()

model.compile(
    optimizer='adam',
    loss='binary_crossentropy',
    metrics=['accuracy', tf.keras.metrics.AUC(name='auc')],
)

epochs = 100

history = model.fit(
    train_ds,
    validation_data=val_ds,
    batch_size=batch_size,
    epochs=epochs,
)

Model: "sequential_3"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
rescaling_3 (Rescaling)      (None, 128, 128, 3)       0         
_________________________________________________________________
conv2d_6 (Conv2D)            (None, 128, 128, 16)      448       
_________________________________________________________________
max_pooling2d_6 (MaxPooling2 (None, 64, 64, 16)        0         
_________________________________________________________________
batch_normalization_2 (Batch (None, 64, 64, 16)        64        
_________________________________________________________________
conv2d_7 (Conv2D)            (None, 64, 64, 32)        4640      
_________________________________________________________________
max_pooling2d_7 (MaxPooling2 (None, 32, 32, 32)        0         
_________________________________________________________________
batch_normalization_3 (Batch (None, 32, 32, 32)        128       
_________________________________________________________________
flatten_3 (Flatten)          (None, 32768)             0         
_________________________________________________________________
dense_6 (Dense)              (None, 128)               4194432   
_________________________________________________________________
dense_7 (Dense)              (None, 1)                 129       
=================================================================
Total params: 4,199,841
Trainable params: 4,199,745
Non-trainable params: 96
_________________________________________________________________
Epoch 1/100
9/9 [==============================] - 1s 62ms/step - loss: 6.4882 - accuracy: 0.4510 - auc: 0.4840 - val_loss: 1.9348 - val_accuracy: 0.5493 - val_auc: 0.5156
Epoch 2/100
9/9 [==============================] - 0s 13ms/step - loss: 1.2934 - accuracy: 0.6617 - auc: 0.7530 - val_loss: 1.3784 - val_accuracy: 0.5070 - val_auc: 0.5561
Epoch 3/100
9/9 [==============================] - 0s 14ms/step - loss: 0.5160 - accuracy: 0.7101 - auc: 0.8111 - val_loss: 0.6844 - val_accuracy: 0.6056 - val_auc: 0.6306
Epoch 4/100
9/9 [==============================] - 0s 13ms/step - loss: 0.4934 - accuracy: 0.6726 - auc: 0.8217 - val_loss: 0.6810 - val_accuracy: 0.5493 - val_auc: 0.5797
Epoch 5/100
9/9 [==============================] - 0s 13ms/step - loss: 0.4005 - accuracy: 0.7577 - auc: 0.8710 - val_loss: 0.6882 - val_accuracy: 0.5352 - val_auc: 0.5857
Epoch 6/100
9/9 [==============================] - 0s 13ms/step - loss: 0.3817 - accuracy: 0.7559 - auc: 0.8725 - val_loss: 0.6850 - val_accuracy: 0.5352 - val_auc: 0.5745
Epoch 7/100
9/9 [==============================] - 0s 12ms/step - loss: 0.3068 - accuracy: 0.7923 - auc: 0.9447 - val_loss: 0.6884 - val_accuracy: 0.4930 - val_auc: 0.5373
Epoch 8/100
9/9 [==============================] - 0s 13ms/step - loss: 0.2817 - accuracy: 0.8721 - auc: 0.9550 - val_loss: 0.6808 - val_accuracy: 0.5352 - val_auc: 0.5641
Epoch 9/100
9/9 [==============================] - 0s 14ms/step - loss: 0.2294 - accuracy: 0.8989 - auc: 0.9718 - val_loss: 0.6758 - val_accuracy: 0.5634 - val_auc: 0.5869
Epoch 10/100
9/9 [==============================] - 0s 13ms/step - loss: 0.2111 - accuracy: 0.9135 - auc: 0.9855 - val_loss: 0.6944 - val_accuracy: 0.5493 - val_auc: 0.5741
Epoch 11/100
9/9 [==============================] - 0s 14ms/step - loss: 0.2111 - accuracy: 0.9516 - auc: 0.9892 - val_loss: 0.6975 - val_accuracy: 0.4930 - val_auc: 0.5252
Epoch 12/100
9/9 [==============================] - 0s 13ms/step - loss: 0.1745 - accuracy: 0.9067 - auc: 0.9961 - val_loss: 0.6969 - val_accuracy: 0.5915 - val_auc: 0.5893
Epoch 13/100
9/9 [==============================] - 0s 13ms/step - loss: 0.1692 - accuracy: 0.9567 - auc: 0.9972 - val_loss: 0.7061 - val_accuracy: 0.5211 - val_auc: 0.5853
Epoch 14/100
9/9 [==============================] - 0s 14ms/step - loss: 0.1435 - accuracy: 0.9802 - auc: 0.9977 - val_loss: 0.6635 - val_accuracy: 0.5775 - val_auc: 0.6450
Epoch 15/100
9/9 [==============================] - 0s 13ms/step - loss: 0.1055 - accuracy: 0.9937 - auc: 0.9992 - val_loss: 0.6587 - val_accuracy: 0.5634 - val_auc: 0.6430
Epoch 16/100
9/9 [==============================] - 0s 13ms/step - loss: 0.0767 - accuracy: 0.9825 - auc: 0.9991 - val_loss: 0.6699 - val_accuracy: 0.5352 - val_auc: 0.6538
Epoch 17/100
9/9 [==============================] - 0s 13ms/step - loss: 0.0476 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 0.6280 - val_accuracy: 0.6338 - val_auc: 0.6883
Epoch 18/100
9/9 [==============================] - 0s 13ms/step - loss: 0.0417 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 0.7108 - val_accuracy: 0.5352 - val_auc: 0.6571
Epoch 19/100
9/9 [==============================] - 0s 15ms/step - loss: 0.0201 - accuracy: 0.9989 - auc: 1.0000 - val_loss: 0.6255 - val_accuracy: 0.6479 - val_auc: 0.6859
Epoch 20/100
9/9 [==============================] - 0s 14ms/step - loss: 0.0174 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 0.6548 - val_accuracy: 0.6197 - val_auc: 0.7071
Epoch 21/100
9/9 [==============================] - 0s 14ms/step - loss: 0.0112 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 0.5995 - val_accuracy: 0.7042 - val_auc: 0.7240
Epoch 22/100
9/9 [==============================] - 0s 13ms/step - loss: 0.0109 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 0.6059 - val_accuracy: 0.6901 - val_auc: 0.7216
Epoch 23/100
9/9 [==============================] - 0s 13ms/step - loss: 0.0079 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 0.5988 - val_accuracy: 0.7042 - val_auc: 0.7308
Epoch 24/100
9/9 [==============================] - 0s 14ms/step - loss: 0.0056 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 0.5986 - val_accuracy: 0.7042 - val_auc: 0.7448
Epoch 25/100
9/9 [==============================] - 0s 14ms/step - loss: 0.0048 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 0.6026 - val_accuracy: 0.7042 - val_auc: 0.7412
Epoch 26/100
9/9 [==============================] - 0s 13ms/step - loss: 0.0044 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 0.6116 - val_accuracy: 0.7042 - val_auc: 0.7444
Epoch 27/100
9/9 [==============================] - 0s 14ms/step - loss: 0.0035 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 0.6074 - val_accuracy: 0.7042 - val_auc: 0.7464
Epoch 28/100
9/9 [==============================] - 0s 13ms/step - loss: 0.0051 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 0.6381 - val_accuracy: 0.6761 - val_auc: 0.7384
Epoch 29/100
9/9 [==============================] - 0s 14ms/step - loss: 0.0042 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 0.6371 - val_accuracy: 0.7183 - val_auc: 0.7364
Epoch 30/100
9/9 [==============================] - 0s 14ms/step - loss: 0.0032 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 0.6653 - val_accuracy: 0.6901 - val_auc: 0.7328
Epoch 31/100
9/9 [==============================] - 0s 13ms/step - loss: 0.0037 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 0.7009 - val_accuracy: 0.6620 - val_auc: 0.7284
Epoch 32/100
9/9 [==============================] - 0s 13ms/step - loss: 0.0026 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 0.7238 - val_accuracy: 0.7324 - val_auc: 0.7300
Epoch 33/100
9/9 [==============================] - 0s 13ms/step - loss: 0.0024 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 0.7622 - val_accuracy: 0.7324 - val_auc: 0.7288
Epoch 34/100
9/9 [==============================] - 0s 13ms/step - loss: 0.0021 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 0.8063 - val_accuracy: 0.7042 - val_auc: 0.7284
Epoch 35/100
9/9 [==============================] - 0s 13ms/step - loss: 0.0027 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 0.8437 - val_accuracy: 0.7042 - val_auc: 0.7320
Epoch 36/100
9/9 [==============================] - 0s 13ms/step - loss: 0.0021 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 0.8883 - val_accuracy: 0.6761 - val_auc: 0.7256
Epoch 37/100
9/9 [==============================] - 0s 13ms/step - loss: 0.0021 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 0.9376 - val_accuracy: 0.6620 - val_auc: 0.7220
Epoch 38/100
9/9 [==============================] - 0s 12ms/step - loss: 0.0020 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 0.9981 - val_accuracy: 0.6479 - val_auc: 0.7228
Epoch 39/100
9/9 [==============================] - 0s 13ms/step - loss: 0.0015 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 1.0560 - val_accuracy: 0.6620 - val_auc: 0.7115
Epoch 40/100
9/9 [==============================] - 0s 14ms/step - loss: 0.0018 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 1.1138 - val_accuracy: 0.6338 - val_auc: 0.7091
Epoch 41/100
9/9 [==============================] - 0s 13ms/step - loss: 0.0014 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 1.1678 - val_accuracy: 0.6479 - val_auc: 0.7111
Epoch 42/100
9/9 [==============================] - 0s 13ms/step - loss: 0.0014 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 1.2324 - val_accuracy: 0.6620 - val_auc: 0.7099
Epoch 43/100
9/9 [==============================] - 0s 30ms/step - loss: 0.0014 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 1.2949 - val_accuracy: 0.6901 - val_auc: 0.7163
Epoch 44/100
9/9 [==============================] - 0s 15ms/step - loss: 0.0015 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 1.3601 - val_accuracy: 0.6761 - val_auc: 0.7183
Epoch 45/100
9/9 [==============================] - 0s 13ms/step - loss: 0.0012 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 1.4324 - val_accuracy: 0.6761 - val_auc: 0.7212
Epoch 46/100
9/9 [==============================] - 0s 14ms/step - loss: 0.0014 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 1.4887 - val_accuracy: 0.6761 - val_auc: 0.7248
Epoch 47/100
9/9 [==============================] - 0s 13ms/step - loss: 0.0012 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 1.5575 - val_accuracy: 0.6761 - val_auc: 0.7284
Epoch 48/100
9/9 [==============================] - 0s 15ms/step - loss: 0.0012 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 1.6339 - val_accuracy: 0.6761 - val_auc: 0.7192
Epoch 49/100
9/9 [==============================] - 0s 14ms/step - loss: 0.0013 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 1.7035 - val_accuracy: 0.6761 - val_auc: 0.7196
Epoch 50/100
9/9 [==============================] - 0s 13ms/step - loss: 0.0013 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 1.7717 - val_accuracy: 0.6901 - val_auc: 0.7248
Epoch 51/100
9/9 [==============================] - 0s 14ms/step - loss: 9.3228e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 1.8471 - val_accuracy: 0.6761 - val_auc: 0.7212
Epoch 52/100
9/9 [==============================] - 0s 13ms/step - loss: 0.0010 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 1.9138 - val_accuracy: 0.6901 - val_auc: 0.7232
Epoch 53/100
9/9 [==============================] - 0s 14ms/step - loss: 0.0010 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 1.9862 - val_accuracy: 0.7183 - val_auc: 0.7272
Epoch 54/100
9/9 [==============================] - 0s 15ms/step - loss: 0.0012 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 2.0603 - val_accuracy: 0.7042 - val_auc: 0.7284
Epoch 55/100
9/9 [==============================] - 0s 14ms/step - loss: 9.8786e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 2.1399 - val_accuracy: 0.6901 - val_auc: 0.7364
Epoch 56/100
9/9 [==============================] - 0s 14ms/step - loss: 8.5569e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 2.2179 - val_accuracy: 0.7324 - val_auc: 0.7304
Epoch 57/100
9/9 [==============================] - 0s 14ms/step - loss: 8.6896e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 2.3019 - val_accuracy: 0.7324 - val_auc: 0.7328
Epoch 58/100
9/9 [==============================] - 0s 14ms/step - loss: 9.9618e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 2.3754 - val_accuracy: 0.7324 - val_auc: 0.7344
Epoch 59/100
9/9 [==============================] - 0s 14ms/step - loss: 7.9414e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 2.4516 - val_accuracy: 0.7042 - val_auc: 0.7356
Epoch 60/100
9/9 [==============================] - 0s 13ms/step - loss: 6.8582e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 2.5276 - val_accuracy: 0.6901 - val_auc: 0.7368
Epoch 61/100
9/9 [==============================] - 0s 13ms/step - loss: 6.5435e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 2.6030 - val_accuracy: 0.6901 - val_auc: 0.7380
Epoch 62/100
9/9 [==============================] - 0s 14ms/step - loss: 6.8066e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 2.6829 - val_accuracy: 0.6901 - val_auc: 0.7372
Epoch 63/100
9/9 [==============================] - 0s 14ms/step - loss: 7.7666e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 2.7543 - val_accuracy: 0.7042 - val_auc: 0.7396
Epoch 64/100
9/9 [==============================] - 0s 14ms/step - loss: 6.6361e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 2.8264 - val_accuracy: 0.6901 - val_auc: 0.7448
Epoch 65/100
9/9 [==============================] - 0s 14ms/step - loss: 6.9111e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 2.8996 - val_accuracy: 0.6901 - val_auc: 0.7448
Epoch 66/100
9/9 [==============================] - 0s 14ms/step - loss: 5.7020e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 2.9771 - val_accuracy: 0.6901 - val_auc: 0.7460
Epoch 67/100
9/9 [==============================] - 0s 14ms/step - loss: 7.0239e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 3.0475 - val_accuracy: 0.6901 - val_auc: 0.7432
Epoch 68/100
9/9 [==============================] - 0s 13ms/step - loss: 5.3068e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 3.1167 - val_accuracy: 0.6901 - val_auc: 0.7400
Epoch 69/100
9/9 [==============================] - 0s 15ms/step - loss: 4.9762e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 3.1823 - val_accuracy: 0.7042 - val_auc: 0.7420
Epoch 70/100
9/9 [==============================] - 0s 13ms/step - loss: 5.9091e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 3.2489 - val_accuracy: 0.7042 - val_auc: 0.7424
Epoch 71/100
9/9 [==============================] - 0s 13ms/step - loss: 5.4181e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 3.3221 - val_accuracy: 0.6901 - val_auc: 0.7308
Epoch 72/100
9/9 [==============================] - 0s 14ms/step - loss: 5.9010e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 3.3755 - val_accuracy: 0.7042 - val_auc: 0.7332
Epoch 73/100
9/9 [==============================] - 0s 13ms/step - loss: 4.3831e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 3.4373 - val_accuracy: 0.6901 - val_auc: 0.7312
Epoch 74/100
9/9 [==============================] - 0s 15ms/step - loss: 4.8843e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 3.4985 - val_accuracy: 0.6901 - val_auc: 0.7284
Epoch 75/100
9/9 [==============================] - 0s 13ms/step - loss: 5.2265e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 3.5523 - val_accuracy: 0.6901 - val_auc: 0.7252
Epoch 76/100
9/9 [==============================] - 0s 13ms/step - loss: 4.6827e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 3.6091 - val_accuracy: 0.6901 - val_auc: 0.7276
Epoch 77/100
9/9 [==============================] - 0s 14ms/step - loss: 4.3389e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 3.6519 - val_accuracy: 0.6901 - val_auc: 0.7268
Epoch 78/100
9/9 [==============================] - 0s 13ms/step - loss: 4.9754e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 3.6938 - val_accuracy: 0.6901 - val_auc: 0.7304
Epoch 79/100
9/9 [==============================] - 0s 14ms/step - loss: 4.0575e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 3.7425 - val_accuracy: 0.6901 - val_auc: 0.7296
Epoch 80/100
9/9 [==============================] - 0s 13ms/step - loss: 4.1413e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 3.7826 - val_accuracy: 0.6901 - val_auc: 0.7244
Epoch 81/100
9/9 [==============================] - 0s 13ms/step - loss: 3.9489e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 3.8244 - val_accuracy: 0.6901 - val_auc: 0.7171
Epoch 82/100
9/9 [==============================] - 0s 15ms/step - loss: 4.6814e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 3.8579 - val_accuracy: 0.6901 - val_auc: 0.7171
Epoch 83/100
9/9 [==============================] - 0s 13ms/step - loss: 3.8877e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 3.8865 - val_accuracy: 0.7042 - val_auc: 0.7171
Epoch 84/100
9/9 [==============================] - 0s 14ms/step - loss: 4.4917e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 3.9212 - val_accuracy: 0.6901 - val_auc: 0.7159
Epoch 85/100
9/9 [==============================] - 0s 14ms/step - loss: 3.3383e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 3.9539 - val_accuracy: 0.6901 - val_auc: 0.7192
Epoch 86/100
9/9 [==============================] - 0s 14ms/step - loss: 3.9012e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 3.9770 - val_accuracy: 0.6901 - val_auc: 0.7196
Epoch 87/100
9/9 [==============================] - 0s 13ms/step - loss: 3.5120e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 4.0039 - val_accuracy: 0.6901 - val_auc: 0.7196
Epoch 88/100
9/9 [==============================] - 0s 14ms/step - loss: 3.9191e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 4.0245 - val_accuracy: 0.6901 - val_auc: 0.7192
Epoch 89/100
9/9 [==============================] - 0s 15ms/step - loss: 3.9131e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 4.0488 - val_accuracy: 0.6901 - val_auc: 0.7091
Epoch 90/100
9/9 [==============================] - 0s 30ms/step - loss: 3.6262e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 4.0725 - val_accuracy: 0.6901 - val_auc: 0.7079
Epoch 91/100
9/9 [==============================] - 0s 13ms/step - loss: 2.9462e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 4.0921 - val_accuracy: 0.6901 - val_auc: 0.7079
Epoch 92/100
9/9 [==============================] - 0s 15ms/step - loss: 2.6391e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 4.1123 - val_accuracy: 0.6901 - val_auc: 0.7067
Epoch 93/100
9/9 [==============================] - 0s 14ms/step - loss: 2.7211e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 4.1287 - val_accuracy: 0.6901 - val_auc: 0.7071
Epoch 94/100
9/9 [==============================] - 0s 14ms/step - loss: 3.3072e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 4.1432 - val_accuracy: 0.6901 - val_auc: 0.7071
Epoch 95/100
9/9 [==============================] - 0s 14ms/step - loss: 2.9676e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 4.1621 - val_accuracy: 0.6901 - val_auc: 0.7059
Epoch 96/100
9/9 [==============================] - 0s 13ms/step - loss: 3.0123e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 4.1733 - val_accuracy: 0.6901 - val_auc: 0.7063
Epoch 97/100
9/9 [==============================] - 0s 14ms/step - loss: 3.3341e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 4.1815 - val_accuracy: 0.6901 - val_auc: 0.7063
Epoch 98/100
9/9 [==============================] - 0s 14ms/step - loss: 2.7530e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 4.2012 - val_accuracy: 0.6901 - val_auc: 0.7055
Epoch 99/100
9/9 [==============================] - 0s 14ms/step - loss: 2.5506e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 4.2160 - val_accuracy: 0.7042 - val_auc: 0.7051
Epoch 100/100
9/9 [==============================] - 0s 14ms/step - loss: 2.6284e-04 - accuracy: 1.0000 - auc: 1.0000 - val_loss: 4.2271 - val_accuracy: 0.7042 - val_auc: 0.7047

結果

plt.figure(figsize=(20, 10))

epochs_range = range(1, epochs+1)
train_loss = history.history['loss']
val_loss = history.history['val_loss']
train_auc = history.history['auc']
val_auc = history.history['val_auc']

plt.subplot(1, 2, 1)
plt.plot(epochs_range, train_loss, label='Training Loss')
plt.plot(epochs_range, val_loss, label='Validation Loss')

plt.xlabel('Epochs')
plt.ylabel('Loss')
plt.title('Training and Validation Loss')
plt.legend()

plt.subplot(1, 2, 2)
plt.plot(epochs_range, train_auc, label='Training AUC')
plt.plot(epochs_range, val_auc, label='Validation AUC')

plt.xlabel('Epochs')
plt.ylabel('AUC')
plt.title('Training and Validation AUC')
plt.legend()

plt.show()

model.evaluate(test_ds)

3/3 [==============================] - 0s 5ms/step - loss: 1.5354 - accuracy: 0.7011 - auc: 0.7565





[1.5353684425354004, 0.7011494040489197, 0.7565360069274902]

np.argmin(val_loss)

23

np.argmax(val_auc)

26