Created
April 15, 2021 09:00
-
-
Save gildniy/5b88cd51f7d66c755464f227c56b9878 to your computer and use it in GitHub Desktop.
Classify structured data using Keras Preprocessing Layers
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
import pandas as pd | |
import tensorflow as tf | |
from sklearn.model_selection import train_test_split | |
from tensorflow.keras.layers.experimental.preprocessing import \ | |
Normalization, CategoryEncoding, IntegerLookup | |
dataset_url = 'https://storage.googleapis.com/kaggle-data-sets/1226038/2047221/bundle/archive.zip?X-Goog-Algorithm=GOOG4-RSA-SHA256&X-Goog-Credential=gcp-kaggle-com%40kaggle-161607.iam.gserviceaccount.com%2F20210414%2Fauto%2Fstorage%2Fgoog4_request&X-Goog-Date=20210414T143749Z&X-Goog-Expires=259199&X-Goog-SignedHeaders=host&X-Goog-Signature=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' | |
tf.keras.utils.get_file('archive.zip', dataset_url, extract=True, | |
cache_dir='.') | |
dataframe = pd.read_csv('datasets/heart.csv') | |
print(dataframe.shape) # (303, 14) | |
print(dataframe.head()) | |
train_df, val_df = train_test_split(dataframe, test_size=0.2) | |
train_df, pred_df = train_test_split(train_df, test_size=0.01) | |
def df_to_dataset(df, predictor, shuffle=True, batch_size=32): | |
df = df.copy() | |
labels = df.pop(predictor) | |
ds = tf.data.Dataset.from_tensor_slices((dict(df), labels)) | |
if shuffle: | |
ds = ds.shuffle(buffer_size=len(df)) | |
ds = ds.batch(batch_size) | |
return ds | |
train_ds = df_to_dataset(train_df, 'output', batch_size=25) | |
val_ds = df_to_dataset(val_df, 'output', shuffle=False, batch_size=25) | |
def get_normalization_layer(name, dataset): | |
normalizer = Normalization() | |
feature_ds = dataset.map(lambda x, y: x[name]) | |
normalizer.adapt(feature_ds) | |
return normalizer | |
def get_category_encoding_layer(name, dataset, max_tokens=None): | |
index = IntegerLookup(max_tokens=max_tokens) | |
feature_ds = dataset.map(lambda x, y: x[name]) | |
index.adapt(feature_ds) | |
encoder = CategoryEncoding(num_tokens=index.vocabulary_size()) | |
feature_ds = feature_ds.map(index) | |
encoder.adapt(feature_ds) | |
return lambda feature: encoder(index(feature)) | |
all_inputs = [] | |
encoded_features = [] | |
numeric_cols = ['age', 'trtbps', 'chol', 'thalachh', 'oldpeak', 'slp'] | |
for header in numeric_cols: | |
numeric_col = tf.keras.Input(shape=(1,), name=header) | |
all_inputs.append(numeric_col) | |
normalization_layer = get_normalization_layer(header, train_ds) | |
encoded_numeric_col = normalization_layer(numeric_col) | |
encoded_features.append(encoded_numeric_col) | |
categorical_cols = ['sex', 'cp', 'fbs', 'restecg', 'exng', 'caa', 'thall'] | |
for header in categorical_cols: | |
categorical_col = tf.keras.Input(shape=(1,), name=header, dtype='int64') | |
all_inputs.append(categorical_col) | |
encoding_layer = get_category_encoding_layer( | |
header, | |
train_ds, | |
max_tokens=5 | |
) | |
encoded_categorical_col = encoding_layer(categorical_col) | |
encoded_features.append(encoded_categorical_col) | |
def build_model(n_units): | |
all_features = tf.keras.layers.concatenate(encoded_features) | |
x = tf.keras.layers.Dense(n_units, activation="relu")(all_features) | |
x = tf.keras.layers.Dropout(0.5)(x) | |
output = tf.keras.layers.Dense(1, activation='sigmoid')(x) | |
model = tf.keras.Model(all_inputs, output) | |
model.compile(optimizer='adam', | |
loss=tf.keras.losses.BinaryCrossentropy(), | |
metrics=["accuracy"]) | |
return model | |
model = build_model(32) | |
tf.keras.utils.plot_model(model, show_shapes=True, rankdir="LR") | |
model.summary() | |
model.fit(train_ds, validation_data=val_ds, epochs=50) | |
loss, accuracy = model.evaluate(val_ds) | |
print("\nLoss: ", loss) | |
print("Accuracy: ", accuracy) | |
sample = list(pred_df.to_dict('index').values())[0] | |
input_dict = {name: tf.convert_to_tensor([value]) for name, value in | |
sample.items() if name != 'output'} | |
predictions = model.predict(input_dict) | |
print(f"\nThis particular patient had a {100 * predictions[0][0]:.1f}% " | |
f"probability of having a heart disease, as evaluated by our model.") |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment