# No labels, use only features.
inputs = features
else:
inputs = (features, labels)
# Convert the inputs to a Dataset.
dataset = tf.data.Dataset.from_tensor_slices(inputs)
# Batch the examples
dataset = dataset.batch(batch_size)
# Return the dataset.
return dataset
train = import_and_clean("UNSW-NB15_1.csv")
test = import_and_clean("UNSW-NB15_2.csv")
input_features = []
for key in train.iloc[:, [9, 36]].keys():
input_features.append(tf.feature_column.numeric_column(key=key))
classifier = tf.estimator.DNNClassifier(feature_columns=input_features,
hidden_units=[10, 10],
n_classes=2)
print("Training...")
start = time.time()
classifier.train(input_fn=lambda: train_input_fn(train.iloc[:, [9, 36]], train.
iloc[:, -1], 100),
steps=1000)
import numpy as np
import matplotlib.pyplot as plt
from sklearn.ensemble import ExtraTreesClassifier
from data_preprocessing_unsw import import_and_clean
ds = import_and_clean("UNSW-NB15_3.csv")
model = ExtraTreesClassifier()
model.fit(ds.iloc[:, :-2], ds.iloc[:, -1])
plt.bar(np.arange(47), model.feature_importances_)
plt.xlabel("Features")
plt.ylabel("Importance")
plt.show()
