Dataset and codes for ABPM-VAE vs. FCNN (Ablation Model) Comparison in Predicting Nocturna

2,920 ambulatory blood pressure monitoring (ABPM) records from 2,876 patients at the First Affiliated Hospital of Wannan Medical College, collected between December 2020 and March 2023. The dataset includes time-stamped systolic and diastolic blood pressure, heart rate, activity status, and posture readings, with binary labels for nocturnal systolic and diastolic hypertension. It was used to compare an ABPM-VAE model against a Fully Connected Neural Network (FCNN) for prediction tasks.

Use Cases

• Predicting nocturnal systolic hypertension based on average nighttime systolic blood pressure readings.
• Predicting nocturnal diastolic hypertension based on average nighttime diastolic blood pressure readings.
• Analyzing the relationship between blood pressure, heart rate, and patient activity status over a 24-hour period.
• Comparing the performance of generative models (VAE) against discriminative models (FCNN) on structured clinical time-series data.

Strengths

• Contains 2,920 records from 2,876 patients, providing a substantial clinical cohort.
• Includes detailed per-measurement data: time, systolic/diastolic BP, heart rate, activity status, and posture.
• Clinical labels for two binary outcomes (nocturnal systolic and diastolic hypertension) are explicitly defined and provided.
• Dataset is split into training/validation (90%) and test (10%) sets with a consistent subset generation strategy.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Data reflects a specific geographic and temporal bias, sourced from a single hospital in China between 2020-2023.
• Row count for the primary data files is unknown, which may limit suitability assessment.

Provenance

Source

First Affiliated Hospital of Wannan Medical College

Collection Method

Ambulatory blood pressure monitoring (ABPM) using DMS-ABP2 monitors, with readings every 30 minutes (day) and 60 minutes (night).

Time Range

December 24, 2020 to March 20, 2023

Freshness

Last updated 2026-04-22 18:36:43; freshness should be verified.

Geography

Likely China (based on hospital name)