[Retracted] Software Systems Security Vulnerabilities Management by Exploring the Capabilities of Language Models Using NLP

Algorithm 3

Classification using USE with TensorFlow 1.0.

	Input: security- and nonsecurity-related text with labeling
	Process:
(1)	Data set split to 50% for training, 10% for validation, and 40% for testing
(2)	TF 2.0 eager execution is disabled as the Google has not updated USE model for compatibility with TF 2.0 (tf.compat.v1.disable_eager_execution ())
(3)	USE model to be loaded from TF Hub: embed = hub.Module (module_url, trainable = True)
(4)	USE embedding layer to be built
(5)	Model architecture to be constructed:
model.compile (loss =
“binary_crossentropy,” optimizer = “Adam,”
metrics = [“accuracy”])
(6)	Latest version of Google’s USE that supports TF 2.0 will be utilized in the further phase of the research
(7)	Training and validation to be conducted for 100 epochs and batch size of 128, with an early stopping approach
(8)	Trained model weights are loaded for inference
(9)	Model performance evaluation on the test data set
	Output:
	Accuracy: 92.61%
	Precision: 0.95
	Recall: 0.93
	F1-score: 0.93