 TensorFlow.org에서 보기 |
 Google Colab에서 실행하기 |
 GitHub에서소스 보기 |
 노트북 다운로드하기 |
이 튜토리얼은 pandas 데이터 프레임을 tf.data.Dataset에 로드하는 방법의 예제를 제공합니다.
이 튜토리얼에서는 Cleveland Clinic Foundation for Heart Disease에서 제공하는 작은 데이터세트를 사용합니다. CSV에는 수백 개의 행이 있습니다. 각 행은 환자를 설명하고 각 열은 속성을 설명합니다. 이 정보를 사용하여 이 데이터세트에서 환자가 심장 질환이 있는지 여부를 예측하는 이진 분류 작업입니다.
pandas를 사용하여 데이터 읽기
import pandas as pd
import tensorflow as tf
심장 데이터세트가 포함된 csv 파일을 다운로드합니다.
csv_file = tf.keras.utils.get_file('heart.csv', 'https://storage.googleapis.com/applied-dl/heart.csv')
pandas를 사용하여 csv 파일을 읽습니다.
df = pd.read_csv(csv_file)
df.head()
df.dtypes
age int64 sex int64 cp int64 trestbps int64 chol int64 fbs int64 restecg int64 thalach int64 exang int64 oldpeak float64 slope int64 ca int64 thal object target int64 dtype: object
데이터 프레임의 object thal 열을 이산 숫자 값으로 변환합니다.
df['thal'] = pd.Categorical(df['thal'])
df['thal'] = df.thal.cat.codes
df.head()
tf.data.Dataset를 사용하여 데이터 로드하기
tf.data.Dataset.from_tensor_slices를 사용하여 pandas 데이터 프레임에서 값을 읽습니다.
tf.data.Dataset를 사용할 때의 이점 중 하나는 간단하고 효율적인 데이터 파이프라인을 작성할 수 있다는 것입니다. 자세한 내용은 데이터 로드 가이드를 참조하세요.
target = df.pop('target')
dataset = tf.data.Dataset.from_tensor_slices((df.values, target.values))
for feat, targ in dataset.take(5):
print ('Features: {}, Target: {}'.format(feat, targ))
Features: [ 63. 1. 1. 145. 233. 1. 2. 150. 0. 2.3 3. 0. 2. ], Target: 0 Features: [ 67. 1. 4. 160. 286. 0. 2. 108. 1. 1.5 2. 3. 3. ], Target: 1 Features: [ 67. 1. 4. 120. 229. 0. 2. 129. 1. 2.6 2. 2. 4. ], Target: 0 Features: [ 37. 1. 3. 130. 250. 0. 0. 187. 0. 3.5 3. 0. 3. ], Target: 0 Features: [ 41. 0. 2. 130. 204. 0. 2. 172. 0. 1.4 1. 0. 3. ], Target: 0
pd.Series는 __array__ 프로토콜을 구현하므로 np.array 또는 tf.Tensor를 사용하는 거의 모든 곳에서 투명하게 사용할 수 있습니다.
tf.constant(df['thal'])
<tf.Tensor: shape=(303,), dtype=int8, numpy=
array([2, 3, 4, 3, 3, 3, 3, 3, 4, 4, 2, 3, 2, 4, 4, 3, 4, 3, 3, 3, 3, 3,
3, 4, 4, 3, 3, 3, 3, 4, 3, 4, 3, 4, 3, 3, 4, 2, 4, 3, 4, 3, 4, 4,
2, 3, 3, 4, 3, 3, 4, 3, 3, 3, 4, 3, 3, 3, 3, 3, 3, 4, 4, 3, 3, 4,
4, 2, 3, 3, 4, 3, 4, 3, 3, 4, 4, 3, 3, 4, 4, 3, 3, 3, 3, 4, 4, 4,
3, 3, 4, 3, 4, 4, 3, 4, 3, 3, 3, 4, 3, 4, 4, 3, 3, 4, 4, 4, 4, 4,
3, 3, 3, 3, 4, 3, 4, 3, 4, 4, 3, 3, 2, 4, 4, 2, 3, 3, 4, 4, 3, 4,
3, 3, 4, 2, 4, 4, 3, 4, 3, 3, 3, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4, 4,
4, 3, 3, 3, 4, 3, 4, 3, 4, 3, 3, 3, 3, 3, 3, 3, 4, 3, 3, 3, 3, 3,
3, 3, 3, 3, 3, 3, 3, 4, 4, 3, 3, 3, 3, 3, 3, 3, 3, 4, 3, 4, 3, 2,
4, 4, 3, 3, 3, 3, 3, 3, 4, 3, 3, 3, 3, 3, 2, 2, 4, 3, 4, 2, 4, 3,
3, 4, 3, 3, 3, 3, 4, 3, 4, 3, 4, 2, 2, 4, 3, 4, 3, 2, 4, 3, 3, 2,
4, 4, 4, 4, 3, 0, 3, 3, 3, 3, 1, 4, 3, 3, 3, 4, 3, 4, 3, 3, 3, 4,
3, 3, 4, 4, 4, 4, 3, 3, 4, 3, 4, 3, 4, 4, 3, 4, 4, 3, 4, 4, 3, 3,
3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 3, 2, 4, 4, 4, 4], dtype=int8)>
데이터세트를 섞고 배치 처리합니다.
train_dataset = dataset.shuffle(len(df)).batch(1)
모델 생성 및 훈련
def get_compiled_model():
model = tf.keras.Sequential([
tf.keras.layers.Dense(10, activation='relu'),
tf.keras.layers.Dense(10, activation='relu'),
tf.keras.layers.Dense(1)
])
model.compile(optimizer='adam',
loss=tf.keras.losses.BinaryCrossentropy(from_logits=True),
metrics=['accuracy'])
return model
model = get_compiled_model()
model.fit(train_dataset, epochs=15)
Epoch 1/15
WARNING:tensorflow:Layer dense is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2. The layer has dtype float32 because its dtype defaults to floatx.
If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.
To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.
303/303 [==============================] - 1s 2ms/step - loss: 1.5284 - accuracy: 0.6502
Epoch 2/15
303/303 [==============================] - 1s 2ms/step - loss: 0.6147 - accuracy: 0.7492
Epoch 3/15
303/303 [==============================] - 1s 2ms/step - loss: 0.5727 - accuracy: 0.7195
Epoch 4/15
303/303 [==============================] - 1s 2ms/step - loss: 0.5679 - accuracy: 0.7294
Epoch 5/15
303/303 [==============================] - 1s 2ms/step - loss: 0.5893 - accuracy: 0.7393
Epoch 6/15
303/303 [==============================] - 1s 2ms/step - loss: 0.5484 - accuracy: 0.7327
Epoch 7/15
303/303 [==============================] - 1s 2ms/step - loss: 0.5191 - accuracy: 0.7492
Epoch 8/15
303/303 [==============================] - 1s 2ms/step - loss: 0.5487 - accuracy: 0.7360
Epoch 9/15
303/303 [==============================] - 1s 2ms/step - loss: 0.5523 - accuracy: 0.7624
Epoch 10/15
303/303 [==============================] - 1s 2ms/step - loss: 0.5234 - accuracy: 0.7459
Epoch 11/15
303/303 [==============================] - 1s 2ms/step - loss: 0.5043 - accuracy: 0.7723
Epoch 12/15
303/303 [==============================] - 1s 2ms/step - loss: 0.5106 - accuracy: 0.7492
Epoch 13/15
303/303 [==============================] - 1s 2ms/step - loss: 0.4850 - accuracy: 0.7657
Epoch 14/15
303/303 [==============================] - 1s 2ms/step - loss: 0.4885 - accuracy: 0.7624
Epoch 15/15
303/303 [==============================] - 1s 2ms/step - loss: 0.4640 - accuracy: 0.7954
<tensorflow.python.keras.callbacks.History at 0x7fe8cc0fe518>
특성 열의 대안
사전을 모델에 대한 입력으로 전달하는 것은 tf.keras.layers.Input 레이어의 일치하는 사전을 작성하는 것만큼 간편하며, 함수형 API를 사용하여 필요한 사전 처리를 적용하고 레이어를 쌓습니다. 특성 열의 대안으로 사용할 수 있습니다.
inputs = {key: tf.keras.layers.Input(shape=(), name=key) for key in df.keys()}
x = tf.stack(list(inputs.values()), axis=-1)
x = tf.keras.layers.Dense(10, activation='relu')(x)
output = tf.keras.layers.Dense(1)(x)
model_func = tf.keras.Model(inputs=inputs, outputs=output)
model_func.compile(optimizer='adam',
loss=tf.keras.losses.BinaryCrossentropy(from_logits=True),
metrics=['accuracy'])
tf.data와 함께 사용했을 때 pd.DataFrame의 열 구조를 유지하는 가장 쉬운 방법은 pd.DataFrame을 dict로 변환하고 해당 사전을 조각화하는 것입니다.
dict_slices = tf.data.Dataset.from_tensor_slices((df.to_dict('list'), target.values)).batch(16)
for dict_slice in dict_slices.take(1):
print (dict_slice)
({'age': <tf.Tensor: shape=(16,), dtype=int32, numpy=
array([63, 67, 67, 37, 41, 56, 62, 57, 63, 53, 57, 56, 56, 44, 52, 57],
dtype=int32)>, 'sex': <tf.Tensor: shape=(16,), dtype=int32, numpy=array([1, 1, 1, 1, 0, 1, 0, 0, 1, 1, 1, 0, 1, 1, 1, 1], dtype=int32)>, 'cp': <tf.Tensor: shape=(16,), dtype=int32, numpy=array([1, 4, 4, 3, 2, 2, 4, 4, 4, 4, 4, 2, 3, 2, 3, 3], dtype=int32)>, 'trestbps': <tf.Tensor: shape=(16,), dtype=int32, numpy=
array([145, 160, 120, 130, 130, 120, 140, 120, 130, 140, 140, 140, 130,
120, 172, 150], dtype=int32)>, 'chol': <tf.Tensor: shape=(16,), dtype=int32, numpy=
array([233, 286, 229, 250, 204, 236, 268, 354, 254, 203, 192, 294, 256,
263, 199, 168], dtype=int32)>, 'fbs': <tf.Tensor: shape=(16,), dtype=int32, numpy=array([1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 1, 0], dtype=int32)>, 'restecg': <tf.Tensor: shape=(16,), dtype=int32, numpy=array([2, 2, 2, 0, 2, 0, 2, 0, 2, 2, 0, 2, 2, 0, 0, 0], dtype=int32)>, 'thalach': <tf.Tensor: shape=(16,), dtype=int32, numpy=
array([150, 108, 129, 187, 172, 178, 160, 163, 147, 155, 148, 153, 142,
173, 162, 174], dtype=int32)>, 'exang': <tf.Tensor: shape=(16,), dtype=int32, numpy=array([0, 1, 1, 0, 0, 0, 0, 1, 0, 1, 0, 0, 1, 0, 0, 0], dtype=int32)>, 'oldpeak': <tf.Tensor: shape=(16,), dtype=float32, numpy=
array([2.3, 1.5, 2.6, 3.5, 1.4, 0.8, 3.6, 0.6, 1.4, 3.1, 0.4, 1.3, 0.6,
0. , 0.5, 1.6], dtype=float32)>, 'slope': <tf.Tensor: shape=(16,), dtype=int32, numpy=array([3, 2, 2, 3, 1, 1, 3, 1, 2, 3, 2, 2, 2, 1, 1, 1], dtype=int32)>, 'ca': <tf.Tensor: shape=(16,), dtype=int32, numpy=array([0, 3, 2, 0, 0, 0, 2, 0, 1, 0, 0, 0, 1, 0, 0, 0], dtype=int32)>, 'thal': <tf.Tensor: shape=(16,), dtype=int32, numpy=array([2, 3, 4, 3, 3, 3, 3, 3, 4, 4, 2, 3, 2, 4, 4, 3], dtype=int32)>}, <tf.Tensor: shape=(16,), dtype=int64, numpy=array([0, 1, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 1, 0, 0, 0])>)
model_func.fit(dict_slices, epochs=15)
Epoch 1/15 19/19 [==============================] - 0s 3ms/step - loss: 12.8710 - accuracy: 0.7261 Epoch 2/15 19/19 [==============================] - 0s 3ms/step - loss: 7.1240 - accuracy: 0.7261 Epoch 3/15 19/19 [==============================] - 0s 3ms/step - loss: 3.2908 - accuracy: 0.5611 Epoch 4/15 19/19 [==============================] - 0s 3ms/step - loss: 3.3019 - accuracy: 0.4917 Epoch 5/15 19/19 [==============================] - 0s 3ms/step - loss: 2.8981 - accuracy: 0.5446 Epoch 6/15 19/19 [==============================] - 0s 3ms/step - loss: 2.8607 - accuracy: 0.5050 Epoch 7/15 19/19 [==============================] - 0s 3ms/step - loss: 2.6697 - accuracy: 0.5248 Epoch 8/15 19/19 [==============================] - 0s 3ms/step - loss: 2.5670 - accuracy: 0.5347 Epoch 9/15 19/19 [==============================] - 0s 3ms/step - loss: 2.4257 - accuracy: 0.5380 Epoch 10/15 19/19 [==============================] - 0s 3ms/step - loss: 2.3069 - accuracy: 0.5347 Epoch 11/15 19/19 [==============================] - 0s 3ms/step - loss: 2.1820 - accuracy: 0.5380 Epoch 12/15 19/19 [==============================] - 0s 3ms/step - loss: 2.0643 - accuracy: 0.5512 Epoch 13/15 19/19 [==============================] - 0s 3ms/step - loss: 1.9490 - accuracy: 0.5545 Epoch 14/15 19/19 [==============================] - 0s 3ms/step - loss: 1.8382 - accuracy: 0.5644 Epoch 15/15 19/19 [==============================] - 0s 3ms/step - loss: 1.7317 - accuracy: 0.5611 <tensorflow.python.keras.callbacks.History at 0x7fe8847e7e48>