Synthetic Health Data with Anomalies for 100 Patients

Synthetic health time-series data includes labels for anomalies and Missing Not At Random (MNAR) gaps. The description indicates vitals follow an AR(1) autoregressive process, suggesting realistic temporal dependencies. The dataset is hosted on Kaggle, but author, license, and update details are unknown.

Use Cases

• Training anomaly detection models based on labeled anomalous patterns.
• Benchmarking imputation methods for handling Missing Not At Random (MNAR) gaps.
• Simulating and studying the behavior of autoregressive (AR(1)) vital sign processes.
• Developing and validating time-series forecasting models for synthetic patient vitals.

Strengths

• Includes explicit anomaly labels, enabling supervised learning tasks.
• Simulates realistic data challenges like MNAR gaps and AR(1) temporal correlations.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

Kaggle

Collection Method

Synthetically generated, as indicated by the title and description.

Time Range

Time-series data is described as covering 10-minute intervals per patient.

Freshness

Last update date is unknown; freshness unverified.

Geography

Synthetic data; no geographic coverage.