MeTGAN: Memory Efficient Tabular GAN for High Cardinality Categorical Datasets

Singh, Shreyansh; Kayathwal, Kanishka; Wadhwa, Hardik; Dhama, Gaurav

doi:10.1007/978-3-030-92310-5_60

Cited by 2 publications

(1 citation statement)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It consists of two neural networks that compete with one another: the discriminator, which is trained to distinguish between real and fake data, and the generator, which creates fictitious data. To trick the discriminator, the generator makes adjustments throughout training, which comprises learning to produce fake data that is indistinguishable from the real data [24].…”

Section: Generative Adversarial Network (Gan)mentioning

confidence: 99%

Utilizing Synthetic Tabular Data Method to Improve Heart Attack Prediction Accuracy

Abd-Alhussain,

Obayes,

Al-Shareefi

2023

AJEST

View full text Add to dashboard Cite

Predicting heart attack possibility is a crucial step in saving human life because it is one of the leading causes of death. How improve and/or achieve as accurate as possible prediction value is a challenging task. Though Deep Learning networks, such as Recurrent Neural Networks (RNNs) and Long Short-Term Memory networks (LSTMs), have been widely used to predict medical events, the issue of providing perfect prediction outcomes is still progressing. The training data insufficiency is a major hindrance to these network's performance. Accordingly, to overcome that issue, this research develops a method for heart attack prediction by using the Conditional Tabular Generative Adversarial Network (CTGAN) model. By anticipating and warning the patient that he may be having a heart attack, he to take the appropriate preventative steps, the proposed method in this research can help to reduce deaths from heart attacks and preserve patients' lives. The suggested method uses the CTGAN model to expand the patient's laboratory test results to build a synthetic trained database. Then, the RNN and LSTM are trained on the generated dataset to enhance the prediction accuracy, and both RNN and LSTM are evaluated before and after the database expansion. The experimental results show that the accuracy of RNN and LSTM is increased to 100% and 99% from 80% and 65%, respectively and the precision of RNN and LSTM is increased to 100% for each of them from 68% and 0, respectively. Finally, sensitivity(recall) increased for RNN and LSTM from 77% and 0 to 100% and 98%, respectively.

show abstract

Section: Generative Adversarial Network (Gan)mentioning

confidence: 99%