ECG5000 Electrocardiogram Time Series Classification Dataset

5000 time series recordings of electrocardiogram (ECG) heartbeats, each 140 data points long, sourced from the OpenML platform. The dataset is designed for classification tasks, distinguishing between normal and arrhythmic heart rhythms. The original collection methodology and time period are not specified in the provided metadata.

Use Cases

• Classify heartbeat type (normal vs. arrhythmic) using the 140-point ECG waveform sequence.
• Benchmark time-series classification models like RNNs or CNNs on the fixed-length 140-point signal.
• Analyze morphological patterns within the 140-point ECG trace to identify characteristic features of arrhythmias.
• Develop anomaly detection systems to flag irregular waveforms within the continuous time-series data.

Strengths

• 5000 individual time-series samples provide a substantial base for model training.
• Fixed-length 140-point sequences ensure consistent input dimensions for algorithms.

Limitations

• Specific column names, clinical labels, and patient demographics are unknown.
• The dataset's temporal coverage, collection geography, and potential class imbalance are not documented.
• Lack of metadata on recording conditions or sensor types may limit clinical interpretability.

Provenance

Source

OpenML platform.

Collection Method

null

Time Range

null

Freshness

null

Geography

null