Cardiovascular synthetic tabular data

This dataset is focuses on cardiovascular diseases. It is generated using a hybrid machine learning model that combines diffusion models with Transformers, emphasizing data privacy. The dataset has been meticulously validated for quality and utility, yielding auspicious results.Validation and Metrics:The dataset has undergone rigorous validation processes to ensure quality, utility, and privacy. These validations involved:

Distance to the Closest Record (DCR): The dataset achieved a DCR of 1.2879. The DCR is a metric that measures the distance of the generated data to the closest record in the original dataset. A higher DCR indicates that the synthetic data closely mirrors the real data in terms of statistical properties, making it reliable for further analysis and research.

Membership Inference Attack Accuracy: The dataset scored 0.6780 in this metric. Membership inference attack accuracy measures the likelihood of correctly inferring whether a particular data point was part of the training dataset. An accuracy of 0.6780 suggests that the model maintains a strong level of privacy. It is important to note that a score of 0.5 would indicate random guessing, hence the achieved score demonstrates significantly better privacy protection than random predictions.

Statistical Tests: Comprehensive statistical tests were conducted to compare the synthetic data with real data. These tests ensure that the synthetic data has similar statistical properties and distributions to the original data.

Machine Learning Efficiency: The utility of the dataset was also validated using machine learning models to ensure that the synthetic data is effective for training and can produce reliable predictive models. The results showed that models trained on this dataset performed well, reinforcing the practical utility of the data.

The high DCR value and the membership inference attack accuracy highlight the balance between data utility and privacy, making this dataset an invaluable resource for researchers and practitioners focusing on cardiovascular diseases and machine learning.